High-level synthesis (HLS) design space exploration (DSE) aims to find Pareto-optimal designs but is hindered by slow synthesis evaluations. Existing graph neural network (GNN) surrogates struggle with homogeneous-style graph representations (causing signal over-squashing) and imprecise source-level heuristics for pragma mapping. We propose Chariot, an automated HLS optimization framework. Chariot leverages LLVM-based static analysis for high-fidelity Use-Def chain tracking, modeling HLS designs as semantic-rich heterogeneous graphs that explicitly map directives to true hardware targets. Our framework achieves state-of-the-art QoR prediction, identifying Pareto-optimal solutions with drastically reduced ranking regret while delivering orders-of-magnitude DSE speedup.

Subspace clustering (SC) approximates high-dimensional data as a combination of low-dimensional subspaces, which is suitable for high-dimensional data analysis across various domains including image segmentation and face recognition. Existing SC methods typically obtain the global structure representation solely through the self-representation of the samples, thereby neglecting the intrinsic local connections among the samples. Moreover, due to their inherent framework design, obtaining additional a priori information in unsupervised scenarios presents a significant challenge. To address these limitations, this paper proposes a new method, named Sample-Dependent Subspace Clustering with Elastic Structure Consistency Constraints (SDSC). Firstly, we introduce a new Elastic Structure Consistency Constraints (ESCC) strategy to measure global and local structures elastically. Benefiting from this strategy, SDSC can flexibly explore the structural information within the samples to obtain a comprehensive data representation. By employing the joint regularization term, SDSC can learn effective cluster assignment information directly from the constrained structured data representation, and the cluster assignment information and representation coefficient matrix are smoothly integrated into a unified framework and learn in a mutually reinforcing manner. This learning approach contributes to comprehensive and high-quality clustering results, enhancing the robustness and utility of SDSC. Extensive experiments on several real-world benchmarks and synthetic datasets demonstrate the feasibility and effectiveness of SDSC.

Accurate and scalable indoor localization for unmanned aerial vehicles (UAVs) is essential for Internet of Things (IoT) applications such as autonomous logistics, infrastructure inspection, and emergency response in GPS-denied environments. However, traditional methods often struggle with cost, deployment complexity, and sensitivity to environmental dynamics, limiting their practicality for large-scale IoT scenarios. This paper presents a method in which Channel State Information (CSI) from low-cost IoT sensors enables robust, device-free 3D UAV localization while optimizing accuracy, sensor adaptability, and data efficiency. We propose CiUAV, leveraging CSI captured by ESP32-S3 sensors, with a Robust CSI Signal Enhancement (RCSE) framework integrating Dynamic AGC Compensation (DAC) and Adaptive Noise Suppression and Outlier Removal (ANSOR), alongside a Sensor-in-Sample (SiS) multi-objective optimization model for adaptive multi-sensor fusion. Experimental evaluations in realistic indoor settings achieve a 3D root mean squared error (RMSE) of 0.2659 meters, outperforming baselines by up to 35% in accuracy and 50% in data efficiency. CiUAV offers a lightweight, scalable, and infrastructure-compatible solution for future IoT-enabled UAV systems.

In the framework of the FD (frequent directions) algorithm, we first develop two efficient algorithms for low-rank matrix approximations under the embedding matrices composed of the product of any SpEmb (sparse embedding) matrix and any standard Gaussian matrix, or any SpEmb matrix and any SRHT (subsampled randomized Hadamard transform) matrix. The theoretical results are also achieved based on the bounds of singular values of standard Gaussian matrices and the theoretical results for SpEmb and SRHT matrices. With a given Tucker-rank, we then obtain several efficient FD-based randomized variants of T-HOSVD (the truncated high-order singular value decomposition) and ST-HOSVD (sequentially T-HOSVD), which are two common algorithms for computing the approximate Tucker decomposition of any tensor with a given Tucker-rank. We also consider efficient FD-based randomized algorithms for computing the approximate TT (tensor-train) decomposition of any tensor with a given TT-rank. Finally, we illustrate the efficiency and accuracy of these algorithms using synthetic and real-world matrix (and tensor) data.

This paper focuses on representation learning for dynamic graphs with temporal interactions. A fundamental issue is that both the graph structure and the nodes own their own dynamics, and their blending induces intractable complexity in the temporal evolution over graphs. Drawing inspiration from the recent progress of physical dynamic models in deep neural networks, we propose Graph Neural Controlled Differential Equations (GN-CDEs), a continuous-time framework that jointly models node embeddings and structural dynamics by incorporating a graph enhanced neural network vector field with a time-varying graph path as the control signal. Our framework exhibits several desirable characteristics, including the ability to express dynamics on evolving graphs without piecewise integration, the capability to calibrate trajectories with subsequent data, and robustness to missing observations. Empirical evaluation on a range of dynamic graph representation learning tasks demonstrates the effectiveness of our proposed approach in capturing the complex dynamics of dynamic graphs.

Transformers have achieved great success in natural language processing and computer vision. The core and basic technique of transformers is the self-attention mechanism. The vanilla self-attention mechanism has quadratic complexity, which limits its applications to vision tasks. Most of the existing linear self-attention mechanisms will sacrifice performance to some extent to reduce complexity. In this paper, we propose a novel linear approximation of the vanilla self-attention mechanism named CURSA to achieve both high performance and low complexity at the same time. CURSA is based on the CUR decomposition to decompose the multiplication of large matrices into the multiplication of several small matrices to achieve almost linear complexity. Experiment results of CURSA in image classification tasks, semantic segmentation tasks, object detection tasks, and long-range arena show that it outperforms state-of-the-art self-attention mechanisms with better data efficiency, faster speed, and higher accuracy.

Compatibilities between the hyperedges of two hypergraphs can be represented as a sparse tensor to avoid exponentially increasing computational costs in hypergraph matching. Kd-tree-based approximate nearest neighbor (ANN) methods have been widely adopted to obtain the sparse compatibility tensor and usually need a relatively high density to guarantee greater accuracy without prior knowledge of the correspondences between a pair of feature point sets. For large scale problems, they require exhaustive computations. This work introduces a novel cascaded second and third-order framework for efficient hypergraph matching. Its core is a CUR decomposition-based sparse compatibility tensor generation method. A rough node assignment is calculated first by a CUR-based pairwise matching process that has a lower computational cost in the second order. Using that intermediate assignment as prior knowledge, a compatibility tensor with higher sparsity can be calculated, with a significantly decreased memory footprint by a novel probability relaxation labeling (PRL)-based hypergraph matching algorithm. The term "reliability" was used to describe how the tensor affects the matching performance and a new measurement, the reliability rate, was proposed to quantify the reliability of a sparse compatibility tensor. Experiment results on large-scale synthetic datasets, and widely adopted benchmarks, demonstrated that the proposed framework outperformed existing methods, creating a more than ten times sparser, but more reliable, compatibility tensor. This proposed CUR-based tensor generation method can be integrated into existing hypergraph matching algorithms and will significantly increase their performance with lower computational costs.

Ensuring personnel safety during power station inspections is a critical yet challenging task due to inherent hazards in such environments. Traditional monitoring methods, including wearable devices and video surveillance, suffer from user discomfort, limited visibility, and high deployment costs. To overcome these limitations, this article proposes PowerHAR, a device-free framework for recognizing power inspection activities based on WiFi channel state information (CSI) acquired from custom-designed ESP32 internet of things (IoT) sensors. PowerHAR introduces a spatio-temporal feature distribution-based power operation recognition network, comprising a transformer-based preprocessing module capable of effectively handling variable-length CSI sequences, and a spatio-temporal extraction module that integrates convolutional operations with multihead self-attention mechanisms for comprehensive feature fusion. By leveraging mutual CSI sensing among distributed sensors, PowerHAR provides robust and accurate recognition of power inspection activities without requiring additional hardware infrastructure. Experimental validation demonstrates that PowerHAR significantly surpasses existing baseline methods, confirming its high reliability and practicality in safety-critical industrial scenarios.

Open set domain adaptation (OSDA) aims to transfer classification-oriented knowledge from a labeled source domain to an unlabeled target domain, which faces the challenges from unseen knowledge in open-set scenarios, i.e., unknown classes privileged to the target domain. Existing methods usually identify unknown classes from classifier prediction directly, which are sensitive to the intrinsic clustering structure and cluster numbers of the unknown class data. In this paper, inspired by the sample relation characterization ability of Optimal Transport (OT), we propose a new type of OT method for OSDA, namely, Target-relaxed Optimal Transport (TROT). Compared with existing OT with strict marginal constraints, TROT imposes a single-side relaxation to the mass requirement on the open-set target domain. Theoretically, we prove that such a relaxation can reduce mis-matches between known and unknown classes, which indicates the transport plan of TROT is promising to identify unknown classes. Methodologically, TROT can identify unknown classes adaptively and map the cross-domain shared data with a sparse plan assignment, which improves both the effectiveness and robustness of known class alignment; besides, a graph embedding with multi-cluster structure of unknown classes is designed to learn a discriminative metric space for open-set classification. Empirically, extensive evaluations are conducted on several image datasets, where TROT achieves significant performance improvements compared with existing techniques for visual recognition in open-set scenarios.

Driver distraction remains a leading cause of traffic accidents, posing a critical threat to road safety globally. As intelligent transportation systems evolve, accurate and real-time identification of driver distraction has become essential. However, existing methods struggle to capture both global contextual and fine-grained local features while contending with noisy labels in training datasets. To address these challenges, we propose DSDFormer, a novel framework that integrates the strengths of Transformer and Mamba architectures through a Dual State Domain Attention (DSDA) mechanism, enabling a balance between long-range dependencies and detailed feature extraction for robust driver behavior recognition. Additionally, we introduce Temporal Reasoning Confident Learning (TRCL), an unsupervised approach that refines noisy labels by leveraging spatiotemporal correlations in video sequences. Beyond achieving state-of-the-art results on AUC-V1, AUC-V2, and 100-Driver datasets, the proposed model is deployable in real-time on embedded platforms such as NVIDIA Jetson AGX Orin and Xavier. Extensive experimental results confirm that DSDFormer and TRCL significantly improve both the accuracy and robustness of driver distraction detection, offering a scalable solution to enhance road safety. Our code has been released athttps://github.com/zhangzr23/driver-noises-learning

To address the surge in wireless traffic, multiple-input multiple-output (MIMO) technology has become essential for advancing communication systems. Within MIMO cellular networks, coordinated beamforming (CBF), achieved through the collaborative design of beamformers across multiple base stations (BSs), presents a promising strategy for enhancing network performance. However, in large-scale multi-cell, multi-user MIMO environments, optimizing beamforming coordination remains challenging due to high-dimensional spaces and dynamic conditions. While centralized, optimization-based CBF algorithms provide near-optimal solutions, their dependence on real-time global channel state information (CSI) and high computational complexity renders them impractical for dynamic networks. To overcome these limitations, we introduce a diffusion model-enhanced multi-agent reinforcement learning (MARL) coordinated beamforming framework, termed Diffusion-enhanced MACBF, which enables BSs to determine optimal beamformers independently based on partial observations. Key innovations of this framework include: firstly, a novel limited-information exchange protocol that facilitates effective coordination among BSs with minimal communication overhead; secondly, the introduction of a diffusion model-enhanced Multi-agent Soft Actor-Critic (MASAC) algorithm that enables BSs to efficiently manage spatial and temporal variations in large-scale scenarios; and thirdly, an optimized network architecture combining an Encoder, Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory (Bi-LSTM) layers to accurately map partial observations to optimal actions. Extensive simulations demonstrate the effectiveness and efficiency of the proposed algorithm, achieving an average rate of 7.5502 bps/Hz with up to 10.66% improvement over existing methods while significantly reducing information exchange requirements.

Music, a traditional element in human entertainment, has been extensively studied. We propose a novel method for music generation from human dance motions. We employ Residual Vector Quantization (RVQ) for music feature tokenization, using RVQ indices as the music representation, which reduces the learning complexity. A cross-modal generation model integrating LSTM and attention mechanisms is designed to generate RVQ codebook indices from dance motions. Finally, the indices can be converted back to music waveforms through a music decoder. Experimental results demonstrate the feasibility of generating coherent music that aligns with dance dynamics, providing a new approach for cross-modal entertainment content creation. ©2025 IEEE.

High-Level Synthesis (HLS) empowers the ability to synthesize a customized hardware description from an untimed software description. However, the quality of the generated hardware is affected by the HLS tool. Current state-of-the-art commercial HLS tools adopt static-scheduling-based algorithms, which perform well for the regular designs but suffer performance degradation for the control-dominant designs. Dynamic scheduling, on the other hand, performs well for control flows but loses certain optimizations, like resource sharing and critical path optimizations, resulting in area overhead and frequency drop. In this paper, we propose a source-to-source transformation to generate an equivalent pseudo cycle-accurate model, so that 1) the transformed code runs dynamically based on different control conditions, and 2) the transformed code still fits in the static HLS tool. As future work, this transformation can be integrated into a compiler to automatically optimize the control-dominant designs in the static-scheduling HLS flow.

3D fluorescence microscopy is essential for understanding fundamental life processes through long-term live-cell imaging. However, due to inherent issues in imaging principles, it faces significant challenges including spatially varying noise and anisotropic resolution, where the axial resolution lags behind the lateral resolution up to 4.5 times. Meanwhile, laser power is kept low to maintain cell viability, leading to inaccessible low-noise and high-resolution paired ground truth (GT). To tackle these limitations, a dual Cycle-consistent Diffusion is proposed to effectively mine intra-volume imaging priors within 3D cell volumes in an unsupervised manner, i.e., Volume Tells (VTCD), achieving de-noising and super-resolution (SR) simultaneously. Specifically, a spatially iso-distributed denoiser is designed to exploit the noise distribution consistency between adjacent low-noise and high-noise regions within the 3D cell volume, suppressing the spatially varying noise. Then, in light of the structural consistency of the cell volume, a cross-plane global-propagation SR module propagates high-resolution details from the XY plane into adjacent regions in the XZ and YZ planes, progressively enhancing resolution across the entire 3D cell volume. Experimental results on 10 in vivo cellular dataset demonstrate high improvements in both de-noising and super-resolution, with axial resolution enhanced from ~ 430 nm to ~ 90 nm.

Foundation medical segmentation models, with MedSAM being the most popular, have achieved promising performance across organs and lesions. However, MedSAM still suffers from compromised performance on specific lesions with intricate structures and appearance, as well as bounding box prompt-induced perturbations. Although current test-time adaptation (TTA) methods for medical image segmentation may tackle this issue, partial (e.g., batch normalization) or whole parametric updates restrict their effectiveness due to limited update signals or catastrophic forgetting in large models. Meanwhile, these approaches ignore the computational complexity during adaptation, which is particularly significant for modern foundation models. To this end, our theoretical analyses reveal that directly refining image embeddings is feasible to approach the same goal as parametric updates under the MedSAM architecture, which enables us to realize high computational efficiency and segmentation performance without the risk of catastrophic forgetting. Under this framework, we propose to encourage maximizing factorized conditional probabilities of the posterior prediction probability using a proposed distribution-approximated latent conditional random field loss combined with an entropy minimization loss. Experiments show that we achieve about 3\% Dice score improvements across three datasets while reducing computational complexity by over 7 times.

Current test- or compression-time adaptation image compression (TTA-IC) approaches, which leverage both latent and decoder refinements as a two-step adaptation scheme, have potentially enhanced the rate-distortion (R-D) performance of learned image compression models on cross-domain compression tasks, \textit{e.g.,} from natural to screen content images. However, compared with the emergence of various decoder refinement variants, the latent refinement, as an inseparable ingredient, is barely tailored to cross-domain scenarios. To this end, we are interested in developing an advanced latent refinement method by extending the effective hybrid latent refinement (HLR) method, which is designed for \textit{in-domain} inference improvement but shows noticeable degradation of the rate cost in \textit{cross-domain} tasks. Specifically, we first provide theoretical analyses, in a cue of marginalization approximation from in- to cross-domain scenarios, to uncover that the vanilla HLR suffers from an underlying mismatch between refined Gaussian conditional and hyperprior distributions, leading to deteriorated joint probability approximation of marginal distribution with increased rate consumption. To remedy this issue, we introduce a simple Bayesian approximation-endowed \textit{distribution regularization} to encourage learning a better joint probability approximation in a plug-and-play manner. Extensive experiments on six in- and cross-domain datasets demonstrate that our proposed method not only improves the R-D performance compared with other latent refinement counterparts, but also can be flexibly integrated into existing TTA-IC methods with incremental benefits.

Understanding protein dynamics are essential for deciphering protein functional mechanisms and developing molecular therapies. However, the complex high-dimensional dynamics and interatomic interactions of biological processes pose significant challenge for existing computational techniques. In this paper, we approach this problem for the first time by introducing Deep Signature, a novel computationally tractable framework that characterizes complex dynamics and interatomic interactions based on their evolving trajectories. Specifically, our approach incorporates soft spectral clustering that locally aggregates cooperative dynamics to reduce the size of the system, as well as signature transform that collects iterated integrals to provide a global characterization of the non-smooth interactive dynamics. Theoretical analysis demonstrates that Deep Signature exhibits several desirable properties, including invariance to translation, near invariance to rotation, equivariance to permutation of atomic coordinates, and invariance under time reparameterization. Furthermore, experimental results on three benchmarks of biological processes verify that our approach can achieve superior performance compared to baseline methods.

The commercial success of generative artificial intelligence (GenAI) has driven an exponential surge in demand for real-time inference in Vision Transformer (ViT) applications, including latency-sensitive domains in autonomous driving, medical imaging and computational photography. This paper introduces FastViT, a high-performance and energy-efficient hardware accelerator for emerging kernel function-based linear attention mechanisms. By leveraging cost-efficient multiplication, mixed-precision quantisation and optimised data flow, FastViT improves real-time performance for high-resolution dense prediction tasks. Compared to existing approaches, experiments demonstrate that FastViT achieves higher throughput and energy efficiency while maintaining negligible accuracy degradation and balanced resource allocation. In the future, we will improve its scalability for next-generation hardware equipped with advanced DSP cores.

The attention mechanism is the key to the success of transformers in different machine learning tasks. However, the quadratic complexity with respect to the sequence length of the vanilla softmax-based attention mechanism becomes the major bottleneck for the application of long sequence tasks, such as vision tasks. Although various efficient linear attention mechanisms have been proposed, they need to sacrifice performance to achieve high efficiency. What's more, memory-efficient methods, such as FlashAttention-1-3, still have quadratic computation complexity which can be further improved. In this paper, we propose a novel efficient linear fast attention (ELFATT) mechanism to achieve low memory input/output operations, linear computational complexity, and high performance at the same time. ELFATT offers 4-7x speedups over the vanilla softmax-based attention mechanism in high-resolution vision tasks without losing performance. ELFATT is FlashAttention friendly. Using FlashAttention-2 acceleration, ELFATT still offers 2-3x speedups over the vanilla softmax-based attention mechanism on high-resolution vision tasks without losing performance. Even in some non-vision tasks of long-range arena, ELFATT still achieves leading performance and offers 1.2-2.3x speedups over FlashAttention-2. Even on edge GPUs, ELFATT still offers 1.6x to 2.0x speedups compared to state-of-the-art attention mechanisms in various power modes from 5W to 60W. Furthermore, ELFATT can be used to enhance and accelerate diffusion tasks directly without training.

We have recently witnessed that ''Intelligence" and `''Compression" are the two sides of the same coin, where the language large model (LLM) with unprecedented intelligence is a general-purpose lossless compressor for various data modalities. This attribute is particularly appealing to the lossless image compression community, given the increasing need to compress high-resolution images in the current streaming media era. Consequently, a spontaneous envision emerges: Can the compression performance of the LLM elevate lossless image compression to new heights? However, our findings indicate that the naive application of LLM-based lossless image compressors suffers from a considerable performance gap compared with existing state-of-the-art (SOTA) codecs on common benchmark datasets. In light of this, we are dedicated to fulfilling the unprecedented intelligence (compression) capacity of the LLM for lossless image compression tasks, thereby bridging the gap between theoretical and practical compression performance. Specifically, we propose P -LLM, a next-pixel prediction-based LLM, which integrates various elaborated insights and methodologies, \textit{e.g.,} pixel-level priors, the in-context ability of LLM, and a pixel-level semantic preservation strategy, to enhance the understanding capacity of pixel sequences for better next-pixel predictions. Extensive experiments on benchmark datasets demonstrate that P-LLM can beat SOTA classical and learned codecs.

This paper proposes the Dual-Stage Sparse Attention (DuSA) mechanism for attention acceleration of transformers. In the first stage, DuSA performs intrablock sparse attention to aggregate local inductive biases. In the second stage, DuSA performs interblock sparse attention to obtain long-range dependencies. Both stages have low computational complexity and can be further accelerated by memory acceleration attention mechanisms directly, which makes DuSA faster than some extremely fast attention mechanisms. The dual-stage sparse attention design provides a lower error in approximating vanilla scaled-dot product attention than the basic single-stage sparse attention mechanisms and further advances the basic sparse attention mechanisms to match or even outperform vanilla scaled-dot product attention. Even in some plug and play situations, DuSA can still maintain low performance loss. DuSA can be used in both training and inference acceleration. DuSA achieves leading performance in different benchmarks: long range arena, image classification, semantic segmentation, object detection, text to video generation, and long context understanding, and accelerates models of different sizes.

Learning the underlying dynamics from data with deep neural networks has shown remarkable potential in modeling various complex physical dynamics. However, current approaches are constrained in their ability to make reliable predictions in a specific domain and struggle with generalizing to unseen systems that are governed by the same general dynamics but differ in environmental characteristics. In this work, we formulate a parameter-efficient method, Fourier Neural Simulator for Dynamical Adaptation (FNSDA), that can readily generalize to new dynamics via adaptation in the Fourier space. Specifically, FNSDA identifies the shareable dynamics based on the known environments using an automatic partition in Fourier modes and learns to adjust the modes specific for each new environment by conditioning on low-dimensional latent systematic parameters for efficient generalization. We evaluate our approach on four representative families of dynamic systems, and the results show that FNSDA can achieve superior or competitive generalization performance compared to existing methods with a significantly reduced parameter cost. Our code is available at https://github.com/WonderSeven/FNSDA.

Polynomial multiplication using Number Theoretic Transform (NTT) is crucial in lattice-based post-quantum cryptography (PQC) and fully homomorphic encryption (FHE), with modulusqsignificantly affecting performance. Fermat moduli of the form$2^{2^{n}} + 1$, such as 65537, offer efficiency gains due to simplified modular reduction and powers-of-2 twiddle factors in NTT. While Fermat moduli have been directly applied or explored for incorporation into existing schemes, Fermat NTT-based polynomial multiplication designs remain underexplored in fully exploiting the benefits of Fermat moduli. This work presents a high-radix/mixed-radix NTT architecture tailored for Fermat moduli, which improves the utilization of the powers-of-2 twiddle factors in large transform sizes. In most cases, our design achieves a 30%–85% reduction in DSP area-time product (ATP) and a 70%–100% reduction in BRAM ATP compared to state-of-the-art designs with smaller or equivalent modulus, while maintaining competitive LUT and FF ATP, underscoring the potential of Fermat NTT-based polynomial multipliers in lattice-based cryptography.

Cellular segmentation in fluorescence images is challenging due to the uneven intensity distribution and distinguishable cell morphology. Existing segmentation models consider the changing cell shape and size very few. We propose a novel multi-scale parallel Swin Transformer fusion network (MSPSTF-Net) for cellular segmentation integrating cell morphological information. Multi-scale parallel Swin Transformer (MSPST) module is designed, consisting of 4 parallel branches at different scales. Each branch contains a self-attention block, which is responsible for learning features at a specific scale and capturing scale-specific information. Moreover, a multi-scale parallel feature fusion (MSPFF) module and a global feature fusion (GFF) module are designed to effectively fuse the multi-scale morphological features. We compare the proposed MSPSTF-Net with existing advanced models on three biological cellular datasets in three metrics, F1 score, AJI, and PQ, while the comprehensive results show that MSPSTF-Net has higher segmentation performance and better generalization ability. Compared to the second place, our method MSPSTF-Net achieves an average improvement of 1.091%, 2.268%, and 1.698% across three metrics on three datasets.

Partial domain adaptation (PDA) is a challenging task in real-world machine learning scenarios. It aims to transfer knowledge from a labeled source domain to a related unlabeled target domain, where the support set of the source label distribution subsumes the target one. Previous PDA works managed to correct the label distribution shift by weighting samples in the source domain. However, the simple reweighing technique cannot explore the latent structure and sufficiently use the labeled data, and then models are prone to over-fitting on the source domain. In this work, we propose a novel importance sampling-based shift correction (IS2C) method, where new labeled data are sampled from a built sampling domain, whose label distribution is supposed to be the same as the target domain, to characterize the latent structure and enhance the generalization ability of the model. We provide theoretical guarantees for IS2C by proving that the generalization error can be sufficiently dominated by IS2C. In particular, by implementing sampling with the mixture distribution, the extent of shift between source and sampling domains can be connected to generalization error, which provides an interpretable way to build IS2C. To improve knowledge transfer, an optimal transport-based independence criterion is proposed for conditional distribution alignment, where the computation of the criterion can be adjusted to reduce the complexity from $\mathcal {O}(n^{3})$ to $\mathcal {O}(n^{2})$ in realistic PDA scenarios. Extensive experiments on PDA benchmarks validate the theoretical results and demonstrate the effectiveness of our IS2C over existing methods.

Ensuring safety in both autonomous driving and advanced driver-assistance systems (ADAS) depends critically on the efficient deployment of traffic sign recognition technology. While current methods show effectiveness, they often compromise between speed and accuracy. To address this issue, we present a novel real-time and efficient road sign detection network, YOLO-TS. This network significantly improves performance by optimizing the receptive fields of multi-scale feature maps to align more closely with the size distribution of traffic signs in various datasets. Moreover, our innovative feature-fusion strategy, leveraging the flexibility of Anchor-Free methods, allows for multi-scale object detection on a high-resolution feature map abundant in contextual information, achieving remarkable enhancements in both accuracy and speed. To mitigate the adverse effects of the grid pattern caused by dilated convolutions on the detection of smaller objects, we have devised a unique module that not only mitigates this grid effect but also widens the receptive field to encompass an extensive range of spatial contextual information, thus boosting the efficiency of information usage. Moreover, to address the scarcity of traffic sign datasets, especially under adverse weather conditions, we introduce two novel datasets: Generated-TT100K-weather and CAWTSSS. Extensive evaluations conducted on challenging public benchmarks—including TT100K, CCTSDB2021, and GTSDB—as well as on our proposed datasets, demonstrate that YOLO-TS surpasses current state-of-the-art methods in both accuracy and inference speed. The code, datasets and weights are available athttps://github.com/Heqiang-Huang/YOLO-TS

Abstract Tensor decomposition algorithms are essential for extracting meaningful latent variables and uncovering hidden structures in real-world data tensors. Unlike conventional deterministic tensor decomposition algorithms, randomized methods offer higher efficiency by reducing memory requirements and computational complexity. This paper proposes an efficient hardware architecture for a randomized tensor decomposition implemented on a field-programmable gate array (FPGA) using high-level synthesis (HLS). The proposed architecture integrates random projection, power iteration, and subspace approximation via QR decomposition to achieve low-rank approximation of multidimensional datasets. The proposed architecture utilizes the capabilities of reconfigurable systems to accelerate tensor computation. It includes three central units: (1) tensor times matrix chain (TTMc), (2) tensor unfolding unit, and (3) QR decomposition unit to implement a three-stage algorithm. Experimental results demonstrate that our FPGA design achieves up to 14.56 times speedup compared to the well-implemented tensor decomposition using software library Tensor Toolbox on an Intel i7-9700 CPU. For a large input tensor of size $$512 \times 512 \times 512$$ 512 × 512 × 512 , the proposed design achieves a 5.55 times speedup compared to an Nvidia Tesla T4 GPU. Furthermore, we utilize our hardware-based high-order singular value decomposition (HOSVD) accelerator for two real applications: background subtraction of dynamic video datasets and data compression. In both applications, our proposed design shows high efficiency regarding accuracy and computational time.

Knee cartilage segmentation for Knee Osteoarthritis (OA) diagnosis is challenging due to domain shifts from varying MRI scanning technologies. Existing cross-modality approaches often use paired order matching or style translation techniques to align features. Still, these methods can sacrifice discrimination in less prominent cartilages and overlook critical higher-order correlations and semantic information. To address this issue, we propose a novel framework called Successive Eigen Noise-assisted Mean Teacher Knowledge Distillation (SEN-MTKD) for adapting 2D knee MRI images across different modalities using partially labeled data. Our approach includes the Eigen Low-rank Subspace (ELRS) module, which employs low-rank approximations to generate meaningful pseudo-labels from domain-invariant feature representations progressively. Complementing this, the Successive Eigen Noise (SEN) module introduces advanced data perturbation to enhance discrimination and diversity in small cartilage classes. Additionally, we propose a subspace-based feature distillation loss mechanism (LRBD) to manage variance and leverage rich intermediate representations within the teacher model, ensuring robust feature representation and labeling. Our framework identifies a mutual cross-domain subspace using higher-order structures and lower energy latent features, providing reliable supervision for the student model. Extensive experiments on public and private datasets demonstrate the effectiveness of our method over state-of-the-art benchmarks. The code is available at github.com/AmmarKhawer/SEN-MTKD.

Action Quality Assessment (AQA) aims to evaluate and score human actions in videos accurately. Existing approaches involve extracting features from the input video and implementing regression based on those features. However, representations derived from a single branch often lack the necessary diversity and flexibility to capture the complexity of human actions effectively. This work addresses these limitations by introducing a multi-branch architecture designed to capture a broad spectrum of video dynamics at varying levels of granularity. Specifically, we enhance video representation in the flow-guided branch by integrating optical flow with video features. This combination of multimodal features offers a more comprehensive context of global motion. Meanwhile, the moment-focused branch is tailored to extract frame-specific features, constructing two distinct quality-based representations with different focuses on moments, which achieves adaptive clues aggregation. Furthermore, the detail-aware branch leverages multiscale deep embeddings from a hierarchy convolutional neural network to capture fine-grained spatial information, which is useful when objects have complex spatial changes. Finally, a post-fusion strategy is employed to merge outputs from all branches, contributing to the comprehensive action quality assessment. Experimental evaluations on three benchmark datasets, FineDiving, MTL-AQA, and AQA-7, demonstrate the superiority of our model in providing reliable assessments of action quality.

Recently, the transform-based tensor representation has attracted increasing attention in multimedia data (e.g., images and videos) recovery problems, which consists of two indispensable components, i.e., the transform and the characterization. Previously, the development of transform-based tensor representation has focused mainly on the transform perspective. Although several attempts have considered shallow matrix factorization (e.g., singular value decomposition and nonnegative matrix factorization) for characterizing the frontal slices of the transformed tensor (termed the latent tensor), the faithful characterization perspective has been underexplored. To address this issue, we propose a unifiedDeepTensorRepresentation (DTR) framework by synergistically combining the deep latent generative module and the deep transform module. Especially, the deep latent generative module can faithfully generate the latent tensor as compared with shallow matrix factorization. The new DTR framework not only allows us to better understand the classical shallow representations but also leads us to explore new representations. To examine the representation capability of the proposed DTR, we consider the representative multidimensional data recovery task and suggest an unsupervised DTR-based multidimensional data recovery model. Extensive experiments demonstrate that DTR achieves superior performance compared to the state-of-the-art methods from both quantitative and qualitative aspects, especially for fine detail recovery.

In this article, we develop an online method called IncTSVD to incrementally compute the tensor singular value decomposition (TSVD) of a given sequence of third-order tensors based on the tensor-tensor concept. This can be considered an extension of incremental SVD based on updating matrices to tensors. IncTSVD is suitable for streamed tensor data and where memory resources are limited. Most existing methods to compute TSVD focus on approximating it using randomized or sketching techniques in a batch setting to decrease the storage and computational costs required. The IncTSVD extends the computation of TSVD to streaming by maintaining the basis tensors of previously arrived data and incrementally updating the approximation using the tensor of incoming data. The computational cost and approximation error of the proposed method were analyzed theoretically and through extensive numerical experiments, which included using synthetic and real-world datasets under streaming scenarios. The IncTSVD method was superior to existing deterministic and randomized tensor decompositions (TDs) based on the t-product for computational and storage costs, and had comparable accuracy to the standard TSVD method.

Low-dose computed tomography (LDCT) image reconstruction techniques can reduce patient radiation exposure while maintaining acceptable imaging quality. Deep learning (DL) is widely used in this problem, but the performance of testing data (also known as target domain) is often degraded in clinical scenarios due to the variations that were not encountered in training data (also known as source domain). Unsupervised domain adaptation (UDA) of LDCT reconstruction has been proposed to solve this problem through distribution alignment. However, existing UDA methods fail to explore the usage of uncertainty quantification, which is crucial for reliable intelligent medical systems in clinical scenarios with unexpected variations. Moreover, existing direct alignment for different patients would lead to content mismatch issues. To address these issues, we propose to leverage a probabilistic reconstruction framework to conduct a joint discrepancy minimization between source and target domains in both the latent and image spaces. In the latent space, we devise a Bayesian uncertainty alignment to reduce the epistemic gap between the two domains. This approach reduces the uncertainty level of target domain data, making it more likely to render well-reconstructed results on target domains. In the image space, we propose a sharpness-aware distribution alignment (SDA) to achieve a match of second-order information, which can ensure that the reconstructed images from the target domain have similar sharpness to normal-dose CT (NDCT) images from the source domain. Experimental results on two simulated datasets and one clinical low-dose imaging dataset show that our proposed method outperforms other methods in quantitative and visualized performance.

Loop pipelining is a key optimization in high-level synthesis (HLS), aimed at overlapping the execution of iterations. Static scheduling, dominant in commercial HLS tools, configures the pipeline based on compile-time analysis, proving conservative for designs with irregular control flow and memory access due to imbalanced recurrences. Speculative Loop pipeline (SLP) is a novel concept that addresses the problem by introducing the speculation and recovery mechanism at the source level to improve the throughput. Although proven promising, it has a significant gap from practical application: It requires accurate early-stage modeling of the pipeline configuration for each path, which is unable to obtain with classic HLS scheduling methods because the SLP process itself interferes with the path length. In this work, we made a step forward by proposing a practical SLP framework with accurate path modeling ability through iterative tuning. We further optimize the SLP technology by combining automatic dataflow extraction with speculative source-level transformation to further boost the performance in specific design patterns. Our framework works on the source level and is easy to be plugged into existing downstream HLS tools. Experiment results demonstrate significant performance improvements over commercial HLS tools and better resource trade-offs compared to the state-of-the-art dynamic-scheduling-based solutions.

Optimal transport (OT) is an important methodology to measure distribution discrepancy, which has achieved promising performance in artificial intelligence applications, e.g., unsupervised domain adaptation. However, from the view of transportation, there are still limitations: 1) the local discriminative structures for downstream tasks, e.g., cluster structure for classification, cannot be explicitly admitted by the learned OT plan; 2) the entropy regularization induces a dense OT plan with increasing uncertainty. To tackle these issues, we propose a novel Probability-Polarized OT (PPOT) framework, which can characterize the structure of OT plan explicitly. Specifically, the probability polarization mechanism is proposed to guide the optimization direction of OT plan, which generates a clear margin between similar and dissimilar transport pairs and reduces the uncertainty. Further, a dynamic mechanism for margin is developed by incorporating task-related information into the polarization, which directly captures the intra/inter class correspondence for knowledge transportation. A mathematical understanding for PPOT is provided from the view of gradient, which ensures interpretability. Extensive experiments on several datasets validate the effectiveness and empirical efficiency of PPOT.

The dramatic surge in the utilisation of generative artificial intelligence (GenAI) underscores the need for a secure and efficient mechanism to responsibly manage, use and disseminate multidimensional data generated by artificial intelligence (AI). In this paper, we propose a blockchain-based copyright traceability framework called ring oscillator-singular value decomposition (RO-SVD), which introduces decomposition computing to approximate low-rank matrices generated from hardware entropy sources and establishes an AI-generated content (AIGC) copyright traceability mechanism at the device level. By leveraging the parallelism and reconfigurability of field-programmable gate arrays (FPGAs), our framework can be easily constructed on existing AI -accelerated devices and provide a low-cost solution to emerging copyright issues of AIGC. We developed a hardware-software (HW /SW) co-design prototype based on comprehensive analysis and on-board experiments with multiple AI-applicable FPGAs. Using AI-generated images as a case study, oyr framework demonstrated effectiveness and emphasised customisation, unpredictability, efficiency, manage-ment and reconfigurability. To the best of our knowledge, this is the first practical hardware study discussing and implementing copyright traceability specifically for AI -generated conten t.

To achieve greater accuracy, hypergraph matching algorithms require exponential increases in computational resources. Recent kd-tree-based approximate nearest neighbor (ANN) methods, despite the sparsity of their compatibility tensor, still require exhaustive calculations for large-scale graph matching. This work utilizes CUR tensor decomposition and introduces a novel cascaded second and third-order hypergraph matching framework (CURSOR) for efficient hypergraph matching. A CUR-based second-order graph matching algorithm is used to provide a rough match, and then the core of CURSOR, a fiber-CUR-based tensor generation method, directly calculates entries of the compatibility tensor by leveraging the initial second-order match result. This significantly decreases the time complexity and tensor density. A probability relaxation labeling (PRL)-based matching algorithm, specifically suitable for sparse tensors, is developed. Experiment results on large-scale synthetic datasets and widely-adopted benchmark sets demonstrate the superiority of CURSOR over existing methods. The tensor generation method in CURSOR can be integrated seamlessly into existing hypergraph matching methods to improve their performance and lower their computational costs.

Cellular segmentation is a crucial step in creating cell shape maps and morphological graphs for living embryos from time-lapse 3D fluorescence images (laser confocal). One reliable method for segmenting cell shapes through deep learning networks is to incorporate voxel distance and topology priors to model shapes in topological structures. However, automated and CNN-based segmentation methods often suffer from low signal-to-noise ratios and insufficient training data. Previous works on semantic segmentation have ignored directional distance and topological information. In this paper, we propose a 3D directional and topological transformer named DTTR (Directional distance mapping and Topological learning TRansformer), which uses topology priors to binarization, and demonstrates an effective directional latent space. We use attention calculation on directional distance maps and utilize topological loss and priors, along with an optimized Delaunay-clustering algorithm, to measure voxel predictions in higher dimensional topology space. DTTR outperforms other existing deep learning models and provides a reliable segmented cell instance dataset (22 new living C. elegans embryos) for establishing 4D cellular morphology map.

Continuum robots, characterized by their high flexibility and infinite degrees of freedom (DoFs), have gained prominence in applications such as minimally invasive surgery and hazardous environment exploration. However, the intrinsic complexity of continuum robots requires a significant amount of time for their motion planning, posing a hurdle to their practical implementation. To tackle these challenges, efficient motion planning methods such as Rapidly Exploring Random Trees (RRT) and its variant, RRT*, have been employed. This paper introduces a unique RRT*-based motion control method tailored for continuum robots. Our approach embeds safety constraints derived from the robots’ posture states, facilitating autonomous navigation and obstacle avoidance in rapidly changing environments. Simulation results show efficient trajectory planning amidst multiple dynamic obstacles and provide a robust performance evaluation based on the generated postures. Finally, preliminary tests were conducted on a two-segment cable-driven continuum robot prototype, confirming the effectiveness of the proposed planning approach. This method is versatile and can be adapted and deployed for various types of continuum robots through parameter adjustments.

3D cellular segmentation of fluorescence images of live embryos is a fundamental step in the analysis of embryonic developmental processes. However, existing fully supervised learning methods (CNN-based) to achieve this often suffer from non-robust loss functions and insufficient training data. Previous work on cellular segmentation did not consider topological information and biological constraints. In this paper, we propose a novel semi-supervised method using topological loss and biologically boundary-aware synthetic (labeled and unlabeled mixed) ground truth for 3D cellular segmentation of live embryos. The topological loss function guides the model to extract features in latent space. Semi-supervised learning and synthetic datasets improve the accuracy of inner and outer membrane recognition on a large number of unlabeled images. Experiment results and evaluation on 1472 live embryo images show that our method outperforms existing deep-learning models. This method can be adapted to images of live embryos of other animals.

Co-clustering simultaneously clusters rows and columns, revealing more fine-grained groups. However, existing co-clustering methods suffer from poor scalability and cannot handle large-scale data. This paper presents a novel and scalable co-clustering method designed to uncover intricate patterns in high-dimensional, large-scale datasets. Specifically, we first propose a large matrix partitioning algorithm that partitions a large matrix into smaller submatrices, enabling parallel co-clustering. This method employs a probabilistic model to optimize the configuration of submatrices, balancing the computational efficiency and depth of analysis. Additionally, we propose a hierarchical co-cluster merging algorithm that efficiently identifies and merges co-clusters from these submatrices, enhancing the robustness and reliability of the process. Extensive evaluations validate the effectiveness and efficiency of our method. Experimental results demonstrate a significant reduction in computation time, with an approximate 83% decrease for dense matrices and up to 30% for sparse matrices.

In monocular 3D human pose estimation, target motions are generally stable and continuous, which indicates that joint velocity can provide valuable information for better estimation. Therefore, it is critical to learn the joint motion trajectory and spatio-temporal information from velocity. Previous works have shown that Transformers are effective in capturing the relationship between tokens. However, in practice, only 2D position is available and 3D velocity has not been explicitly used as a model input. To address this challenge, we propose TMT (Two-step Mixed-Training strategy), a transformer-based approach that effectively incorporates 3D velocity into the input vector during training, allowing for better learning of relevant features in the shallow layers. Extensive experiments demonstrate that TMT significantly improves the performance of state-of-the-art models, such as MixSTE, MHFormer, and PoseFomer, on two datasets: Human3.6M and MPI-INF-3DHP. TMT outperforms the state-of-the-art approach by up to 13.8% on the Human3.6M dataset.

MOTIVATION: Automated cell lineage tracing throughout embryogenesis plays a key role in the study of regulatory control of cell fate differentiation, morphogenesis and organogenesis in the development of animals, including nematode Caenorhabditis elegans. However, automated cell lineage tracing suffers from an exponential increase in errors at late embryo because of the dense distribution of cells, relatively low signal-to-noise ratio (SNR) and imbalanced intensity profiles of fluorescence images, which demands a huge amount of human effort to manually correct the errors. The existing image enhancement methods are not sensitive enough to deal with the challenges posed by the crowdedness and low signal-to-noise ratio. An alternative method is urgently needed to assist the existing detection methods in improving their detection and tracing accuracy, thereby reducing the huge burden for manual curation. RESULTS: We developed a new method, termed as DELICATE, that dramatically improves the accuracy of automated cell lineage tracing especially during the stage post 350 cells of C. elegans embryo. DELICATE works by increasing the local SNR and improving the evenness of nuclei fluorescence intensity across cells especially in the late embryos. The method both dramatically reduces the segmentation errors by StarryNite and the time required for manually correcting tracing errors up to 550-cell stage, allowing the generation of accurate cell lineage at large-scale with a user-friendly software/interface. AVAILABILITY AND IMPLEMENTATION: All images and data are available at https://doi.org/10.6084/m9.figshare.26778475.v1. The code and user-friendly software are available at https://github.com/plcx/NucApp-develop.

Abstract In this work, we propose to tackle the problem of domain generalization in the context of insufficient samples. Instead of extracting latent feature embeddings based on deterministic models, we propose to learn a domain-invariant representation based on the probabilistic framework by mapping each data point into probabilistic embeddings. Specifically, we first extend empirical maximum mean discrepancy (MMD) to a novel probabilistic MMD that can measure the discrepancy between mixture distributions (i.e., source domains) consisting of a series of latent distributions rather than latent points. Moreover, instead of imposing the contrastive semantic alignment (CSA) loss based on pairs of latent points, a novel probabilistic CSA loss encourages positive probabilistic embedding pairs to be closer while pulling other negative ones apart. Benefiting from the learned representation captured by probabilistic models, our proposed method can marriage the measurement on the distribution over distributions (i.e., the global perspective alignment) and the distribution-based contrastive semantic alignment (i.e., the local perspective alignment). Extensive experimental results on three challenging medical datasets show the effectiveness of our proposed method in the context of insufficient data compared with state-of-the-art methods.

Generalizable medical image segmentation enables models to generalize to unseen target domains under domain shift issues. Recent progress demonstrates that the shape of the segmentation objective, with its high consistency and robustness across domains, can serve as a reliable regularization to aid the model for better cross-domain performance, where existing methods typically seek a shared framework to render segmentation maps and shape prior concurrently. However, due to the inherent texture and style preference of modern deep neural networks, the edge or silhouette of the extracted shape will inevitably be undermined by those domain-specific texture and style interferences of medical images under domain shifts. To address this limitation, we devise a novel framework with a separation between the shape regularization and the segmentation map. Specifically, we first customize a novel whitening transform-based probabilistic shape regularization extractor namely WT-PSE to suppress undesirable domain-specific texture and style interferences, leading to more robust and high-quality shape representations. Second, we deliver a Wasserstein distance-guided knowledge distillation scheme to help the WT-PSE to achieve more flexible shape extraction during the inference phase. Finally, by incorporating domain knowledge of medical images, we propose a novel instance-domain whitening transform method to facilitate a more stable training process with improved performance. Experiments demonstrate the performance of our proposed method on both multi-domain and single-domain generalization.

Generating choreography from music poses a significant challenge. Conventional dance generation methods are limited by only being able to match specific dance movements to music with corresponding rhythms, restricting the utilization of existing dance sequences. To address this limitation, we propose a method that generates a label, based on a probability distribution function derived from music features, that can be applied to music segments of varying lengths. By using the Kullback-Leibler divergence, we assess the similarity between music segments based on these labels. To ensure adaptability to different musical rhythms, we employ a cubic spline method to represent dance movements. This approach allows us to control the speed of a dance sequence by resampling it, enabling adaptation to varying rhythms based on the tempo of newly input music. To evaluate the effectiveness of our method, we compared the dances generated by our approach with those generated by other neural network-based and conventional methods. Quantitative evaluations demonstrated that our method outperforms these alternatives in terms of dance quality and fidelity.

Facial micro-expressions (MEs) refer to brief spontaneous facial movements that can reveal a person's genuine emotion. They are valuable in lie detection, criminal analysis, and other areas. While deep learning-based ME recognition (MER) methods achieved impressive success, these methods typically require pre-processing using conventional optical flow-based methods to extract facial motions as inputs. To overcome this limitation, we proposed a novel MER framework using self-supervised learning to extract facial motion for ME (SelfME). To the best of our knowledge, this is the first work using an automatically self-learned motion technique for MER. However, the self-supervised motion learning method might suffer from ignoring symmetrical facial actions on the left and right sides of faces when extracting fine features. To address this issue, we developed a symmetric contrastive vision transformer (SCViT) to constrain the learning of similar facial action features for the left and right parts of faces. Experiments were conducted on two benchmark datasets showing that our method achieved state-of-the-art performance, and ablation studies demonstrated the effectiveness of our method.

Recently, successive subspace learning (SSL)-based methods have shown to be effective for the task of visual object classification with mild data desire and mathematically transparent interpretable capability. However, existing SSL-based methods rely heavily on the data-centric subspace representations, leading to potential performance degradation problem in case of the domain shift between the training (a.k.a., source domain) and testing (a.k.a., target domain) data. To address this limitation, we propose an effective successive subspace learning method based on existing SSL-based methods. Specifically, we introduce a novel linear transformation layer to align eigenvectors in SSL module between source and target domains, as such, the discrepancy between source and target domains will be reduced, resulting in better cross-domain performance. The effectiveness of our proposed method is demonstrated on the Office-Caltech-10 and Office-31 benchmark datasets by using features extracted from pre-trained deep neural networks as input.

Many Computer-Generated (CG) images are spreading widely on the Internet, which may deliberately misinform or deceive the public. Therefore, distinguishing CG images from natural photographic (PG) has become a frontier research topic in the field of image forensics. Although many algorithms have been proposed, it is still very challenging to detect CG images generated by the recent cutting-edge generative methods. Besides, most existing algorithms tend to generalize poorly when facing different unseen multimodal generative models. To address this issue, a novel method based on amplified texture differences learning is proposed to tackle this problem. We first design a deep texture enhancement module for discriminative texture amplification. Specifically, a semantic segmentation module is utilized to generate semantic segmentation map for the affine transformation operation guidance, which can be further used to recover the texture in different regions of the input image. Then, the combination of the original image and the high-frequency components of the original and enhanced images are fed into a hybrid neural network equipped with attention mechanisms, which refines intermediate features and facilitates trace exploration in spatial and channel dimensions respectively. By verifying on several commonly used benchmark datasets and a newly constructed dataset11The benchmark is available at https://github.com/191578010/DSGCG. with more realistic and diverse images, the experimental results demonstrate that the proposed approach outperforms some existing methods.

The single-cell RNA sequencing (scRNA-seq) technique begins a new era by revealing gene expression patterns at single-cell resolution, enabling studies of heterogeneity and transcriptome dynamics of complex tissues at single-cell resolution. However, existing large proportion of dropout events may hinder downstream analyses. Thus imputation of dropout events is an important step in analyzing scRNA-seq data. We develop scTSSR2, a new imputation method that combines matrix decomposition with the previously developed two-side sparse self-representation, leading to fast two-side sparse self-representation to impute dropout events in scRNA-seq data. The comparisons of computational speed and memory usage among different imputation methods show that scTSSR2 has distinct advantages in terms of computational speed and memory usage. Comprehensive downstream experiments show that scTSSR2 outperforms the state-of-the-art imputation methods. A user-friendly R package scTSSR2 is developed to denoise the scRNA-seq data to improve the data quality.

Lung cancer is a major cause of cancer deaths worldwide, and has a very low survival rate. Non-small cell lung cancer (NSCLC) is the largest subset of lung cancers, which accounts for about 85% of all cases. It has been well established that a mutation in the epidermal growth factor receptor (EGFR) can lead to lung cancer. EGFR Tyrosine Kinase Inhibitors (TKIs) are developed to target the kinase domain of EGFR. These TKIs produce promising results at the initial stage of therapy, but the efficacy becomes limited due to the development of drug resistance. In this paper, we provide a comprehensive overview of computational methods, for understanding drug resistance mechanisms. The important EGFR mutants and the different generations of EGFR-TKIs, with the survival and response rates are discussed. Next, we evaluate the role of important EGFR parameters in drug resistance mechanism, including structural dynamics, hydrogen bonds, stability, dimerization, binding free energies, and signaling pathways. Personalized drug resistance prediction models, drug response curve, drug synergy, and other data-driven methods are also discussed. Recent advancements in deep learning; such as AlphaFold2, deep generative models, big data analytics, and the applications of statistics and permutation are also highlighted. We explore limitations in the current methodologies, and discuss strategies to overcome them. We believe this review will serve as a reference for researchers; to apply computational techniques for precision medicine, analyzing structures of protein-drug complexes, drug discovery, and understanding the drug response and resistance mechanisms in lung cancer patients.

Covid-19 has become a world pandemic for years. With the appearance of mutations, immune escape has become a problem, reducing the effectiveness of vaccines and antibodies. To reveal the mechanism of immune escape, we analyze the geometrical properties of the receptor-binding domain in the SARS-CoV-2 spike protein, which plays a vital role in the immune reaction. Several important variants are taken as examples, and the wild type model is prepared as a reference. The computational method is applied to simulate the behaviors of the models, and alpha shape algorithm is employed to extract geometrical data of the protein surface. Average moving distance of the surface atoms is used to quantify their activity. Our results show that the mutations changed the properties of the protein. The variants have different distributions of active sites, which may change the specific antigenicity and influence the binding abilities of drugs and antibodies. This study explains the mechanism of immune escape of SARS-CoV-2, and provides a geometrical method to find potential new target sites for the design of drugs and vaccines.

A matrix factorization is to decompose a matrix into a product of smaller matrices. It is widely used in machine learning algorithms. There are many matrix decomposition algorithms, and each has various applications. CUR matrix decomposition is a widely-used factorization tool that has been employed for dimension reduction and pattern recognition in many scientific and engineering applications, such as image processing, text mining, and wireless communications. In this paper we propose an efficient FPGA-based floating-point accelerator using high-level synthesis (HLS) for the CUR decomposition algorithm. Our experiment results demonstrate the better efficiency of our hardware design compared to the optimized CPU-based software solutions. The speedup of our FPGA-based architecture over the optimized software implementation ranges from 2.37 to 16.82 times for different dimensions of the data input matrix. We evaluated our design using large dimension matrices 1024 x 1024 and 2048 x 2048 and the experiment results demonstrated the efficiency of our design in terms of the utilized resources and latency. Finally, we have compared our design with other matrix decomposition algorithms such as SVD and QR decomposition, the experiment results demonstrated that CUR is more efficient than SVD and QR decomposition in terms of latency and required resources.

Convolutional neural networks (CNNs) are usually used as a backbone to design methods in biomedical image segmentation. However, the limitation of receptive field and large number of parameters limit the performance of these methods. In this paper, we propose a graph neural network (GNN) based method named GNN-SEG for the segmentation of brain tissues. Different to conventional CNN based methods, GNN-SEG takes superpixels as basic processing units and uses GNNs to learn the structure of brain tissues. Besides, inspired by the interaction mechanism in biological vision systems, we propose two kinds of interaction modules for feature enhancement and integration. In the experiments, we compared GNN-SEG with state-of-the-art CNN based methods on four datasets of brain magnetic resonance images. The experimental results show the superiority of GNN-SEG.

Facial micro-expression (ME) refers to a brief spontaneous facial movement that can reveal the genuine emotion of a person. The absence of data is a major problem for ME. Thankfully, generative deep neural network models can aid in producing desired samples. In this work, we proposed a deep learning based adaptive dual motion model (ADMM) for generating facial ME samples. A dual motion extraction (DME) module extracts robust motions from two modalities: original color images and edge-based grayscale images, with dual streams. Using edge-based grayscale images can help the method focus on learning subtle movements by eliminating the influences of noises and illumination variants. The motions extracted by the dual streams are fed into an adaptive motion fusion (AMF) module for combing the motions adaptively to generate the dense motion. Our method was trained on the CASME II, SMIC, and SAMM datasets. The evaluation and analysis of the results demonstrated the effectiveness of our method.

The identification of differentially expressed genes between different cell groups is a crucial step in analyzing single-cell RNA-sequencing (scRNA-seq) data. Even though various differential expression analysis methods for scRNA-seq data have been proposed based on different model assumptions and strategies recently, the differentially expressed genes identified by them are quite different from each other, and the performances of them depend on the underlying data structures. In this paper, we propose a new ensemble learning-based differential expression analysis method, scDEA, to produce a more stable and accurate result. scDEA integrates the P-values obtained from 12 individual differential expression analysis methods for each gene using a P-value combination method. Comprehensive experiments show that scDEA outperforms the state-of-the-art individual methods with different experimental settings and evaluation metrics. We expect that scDEA will serve a wide range of users, including biologists, bioinformaticians and data scientists, who need to detect differentially expressed genes in scRNA-seq data.

MOTIVATION: Single-cell RNA sequencing (scRNA-seq) technologies have been testified revolutionary for their promotion on the profiling of single-cell transcriptomes at single-cell resolution. Excess zeros due to various technical noises, called dropouts, will mislead downstream analyses. Therefore, it is crucial to have accurate imputation methods to address the dropout problem. RESULTS: In this article, we develop a new dropout imputation method for scRNA-seq data based on multi-objective optimization. Our method is different from existing ones, which assume that the underlying data has a preconceived structure and impute the dropouts according to the information learned from such structure. We assume that the data combines three types of latent structures, including the horizontal structure (genes are similar to each other), the vertical structure (cells are similar to each other) and the low-rank structure. The combination weights and latent structures are learned using multi-objective optimization. And, the weighted average of the observed data and the imputation results learned from the three types of structures are considered as the final result. Comprehensive downstream experiments show the superiority of our method in terms of recovery of true gene expression profiles, differential expression analysis, cell clustering and cell trajectory inference. AVAILABILITY AND IMPLEMENTATION: The R package is available at https://github.com/Zhangxf-ccnu/scMOO and https://zenodo.org/record/5785195. The codes to reproduce the downstream analyses in this article can be found at https://github.com/Zhangxf-ccnu/scMOO_experiments_codes and https://zenodo.org/record/5786211. The detailed list of data sets used in the present study is represented in Supplementary Table S1 in the Supplementary materials. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Unsupervised domain adaptation is effective in leveraging rich information from a labeled source domain to an unlabeled target domain. Though deep learning and adversarial strategy made a significant breakthrough in the adaptability of features, there are two issues to be further studied. First, hard-assigned pseudo labels on the target domain are arbitrary and error-prone, and direct application of them may destroy the intrinsic data structure. Second, batch-wise training of deep learning limits the characterization of the global structure. In this paper, a Riemannian manifold learning framework is proposed to achieve transferability and discriminability simultaneously. For the first issue, this framework establishes a probabilistic discriminant criterion on the target domain via soft labels. Based on pre-built prototypes, this criterion is extended to a global approximation scheme for the second issue. Manifold metric alignment is adopted to be compatible with the embedding space. The theoretical error bounds of different alignment metrics are derived for constructive guidance. The proposed method can be used to tackle a series of variants of domain adaptation problems, including both vanilla and partial settings. Extensive experiments have been conducted to investigate the method and a comparative study shows the superiority of the discriminative manifold learning framework.

Due to the development of high-throughput technologies for gene analysis, the biclustering method has attracted much attention. However, existing methods have problems with high time and space complexity. This paper proposes a biclustering method, called Row and Column Structure-based Biclustering (RCSBC), with low time and space complexity to find checkerboard patterns within microarray data. First, the paper describes the structure of bicluster by using the structure of rows and columns. Second, the paper chooses the representative rows and columns with two algorithms. Finally, the gene expression data are biclustered on the space spanned by representative rows and columns. To the best of our knowledge, this paper is the first to exploit the relationship between the row/column structure of a gene expression matrix and the structure of biclusters. Both the synthetic datasets and the real-life gene expression datasets are used to validate the effectiveness of our method. It can be seen from the experiment results that the RCSBC outperforms the state-of-the-art algorithms both on clustering accuracy and time/space complexity. This study offers new insights into biclustering the large-scale gene expression data without loading the whole data into memory.

Non-small cell lung cancer with an activating epidermal growth factor receptor (EGFR) mutation responds well to targeted drugs. In most cases, drug resistance appears after about a year. Several studies have been conducted on the kinase domain of EGFR to understand the drug resistance mechanism. Since EGFR is a multi-domain protein, mutation in the kinase domain may affect the other domains as well. In this study, we examine the complete structure of the multi-domain EGFR protein and its mutants. We performed molecular dynamics simulations for wildtype EGFR, EGFR with L858R mutation, and EGFR with L858R and T790M mutations. We applied normal mode analysis and complex network analysis to extract the correlated motions in the domains of EGFR. The normal modes are used to construct the dynamic cross-correlation map (DCCM). Simulation results show different patterns of correlated motions in each domain of EGFR mutants compared to the wildtype. In Domains 1 and 3 of the extracellular region, a small number of weak positively correlated motions are extracted. Domains 2 and 4 show large numbers of both positive and negative motions. However, the negatively correlated motions are stronger in mutant structures compared to the wildtype. In Domain 7, some residues showed a positive correlation around the main diagonal. We also identified different communities, nodes and crucial residues in the domains of the structures, which can be important for the function of EGFR. Moreover, hydrogen bond analysis is performed for the stability analysis. The mutant structures have fewer hydrogen bonds compared to the wildtype. Overall, these findings are useful for understanding the dynamics and communications in EGFR domains.

It is an important task to learn how gene regulatory networks change under different conditions. Several Gaussian graphical model-based methods have been proposed to deal with this task by inferring differential networks from gene expression data. However, most existing methods define the differential networks as the difference of precision matrices, which may include false differential edges caused by the change of conditional variances. In addition, prior information about the condition-specific networks and the differential networks can be obtained from other domains. It is useful to incorporate prior information into differential network analysis. In this study, we propose a new differential network analysis method to address the above challenges. Instead of using the precision matrices, we define the differential networks as the difference of partial correlations, which can exclude the spurious differential edges due to the variants of conditional variances. Furthermore, prior information from multiple hypothesis testing is incorporated using a weighted fused penalty. Simulation studies show that our method outperforms the competing methods. We also apply our method to identify the differential network between luminal A and basal-like subtypes of breast cancers and the differential network between acute myeloid leukemia tumors and normal samples. The hub genes in the differential networks identified by our method carry out important biological functions.

Inferring gene co-expression networks from high-throughput gene expression data is an important task in bioinformatics. Many gene networks often exhibit modular structures. Although several Gaussian graphical model-based methods have been developed to estimate gene co-expression networks by incorporating the modular structural prior, none of them takes into account the modular structures captured by the prior networks (e.g., protein interaction networks). In this study, we propose a novel prior network-dependent gene network inference (pGNI) method to estimate gene co-expression networks by integrating gene expression data and prior protein interaction network data. The underlying modular structure is learned from both sets of data. Through simulation studies, we demonstrate the feasibility and effectiveness of our method. We also apply our method to two real datasets. The modular structures in the networks estimated by our method are biological significant.

Semantic and instance segmentation algorithms are two general yet distinct image segmentation solutions powered by Convolution Neural Network. While semantic segmentation benefits extensively from the end-to-end training strategy, instance segmentation is frequently framed as a multi-stage task, supported by learning-based discrimination and post-process clustering. Independent optimizations on substages instigate the accumulation of segmentation errors. In this work, we propose to embed prior clustering information into an embedding learning framework FCRNet, stimulating the one-stage instance segmentation. FCRNet relieves the complexity of post process by incorporating the number of clustering groups into the embedding space. The superior performance of FCRNet is verified and compared with other methods on the nucleus dataset BBBC006.

Facial micro-expression (FME) refers to a brief spontaneous facial movement that can reveal a person's genius emotion. One challenge in facial micro-expression is the lack of data. Fortunately, generative deep neural network models can assist in the creation of desired images. However, the issues for micro-expressions are the facial variations are too subtle to capture, and the limited training data may make feature extraction difficult. To address these issues, we developed a deep motion retargeting and transfer learning based facial micro-expression generation model (DMT-FMEG). First, to capture subtle variations, we employed a deep motion retargeting (DMR) network that can learn keypoints in an unsupervised manner, estimate motions, and generate desired images. Second, to enhance the feature extraction ability, we applied deep transfer learning (DTL) by borrowing knowledge from macro-expression images. We evaluated our method on three datasets, CASME II, SMIC, and SAMM, and found that it showed satisfactory results on all of them. With the effectiveness of the method, we won the second place in the generation task of the FME 2021 challenge.

The mechanisms controlling biological process, such as the development of disease or cell differentiation, can be investigated by examining changes in the networks of gene dependencies between states in the process. High-throughput experimental methods, like microarray and RNA sequencing, have been widely used to gather gene expression data, which paves the way to infer gene dependencies based on computational methods. However, most differential network analysis methods are designed to deal with fully observed data, but missing values, such as the dropout events in single-cell RNA-sequencing data, are frequent. New methods are needed to take account of these missing values. Moreover, since the changes of gene dependencies may be driven by certain perturbed genes, considering the changes in gene expression levels may promote the identification of gene network rewiring. In this study, a novel weighted differential network estimation (WDNE) model is proposed to handle multi-platform gene expression data with missing values and take account of changes in gene expression levels. Simulation studies demonstrate that WDNE outperforms state-of-the-art differential network estimation methods. When applied WDNE to infer differential gene networks associated with drug resistance in ovarian tumors, cell differentiation and breast tumor heterogeneity, the hub genes in the estimated differential gene networks can provide important insights into the underlying mechanisms. Furthermore, a Matlab toolbox, differential network analysis toolbox, was developed to implement the WDNE model and visualize the estimated differential networks.

Accurately predicting protein-ligand binding affinities can substantially facilitate the drug discovery process, but it remains as a difficult problem. To tackle the challenge, many computational methods have been proposed. Among these methods, free energy-based simulations and machine learning-based scoring functions can potentially provide accurate predictions. In this paper, we review these two classes of methods, following a number of thermodynamic cycles for the free energy-based simulations and a feature-representation taxonomy for the machine learning-based scoring functions. More recent deep learning-based predictions, where a hierarchy of feature representations are generally extracted, are also reviewed. Strengths and weaknesses of the two classes of methods, coupled with future directions for improvements, are comparatively discussed.

MOTIVATION: Clustering analysis in a biological network is to group biological entities into functional modules, thus providing valuable insight into the understanding of complex biological systems. Existing clustering techniques make use of lower-order connectivity patterns at the level of individual biological entities and their connections, but few of them can take into account of higher-order connectivity patterns at the level of small network motifs. RESULTS: Here, we present a novel clustering framework, namely HiSCF, to identify functional modules based on the higher-order structure information available in a biological network. Taking advantage of higher-order Markov stochastic process, HiSCF is able to perform the clustering analysis by exploiting a variety of network motifs. When compared with several state-of-the-art clustering models, HiSCF yields the best performance for two practical clustering applications, i.e. protein complex identification and gene co-expression module detection, in terms of accuracy. The promising performance of HiSCF demonstrates that the consideration of higher-order network motifs gains new insight into the analysis of biological networks, such as the identification of overlapping protein complexes and the inference of new signaling pathways, and also reveals the rich higher-order organizational structures presented in biological networks. AVAILABILITY AND IMPLEMENTATION: HiSCF is available at https://github.com/allenv5/HiSCF. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Differential network analysis is an important tool to investigate the rewiring of gene interactions under different conditions. Several computational methods have been developed to estimate differential networks from gene expression data, but most of them do not consider that gene network rewiring may be driven by the differential expression of individual genes. New differential network analysis methods that simultaneously take account of the changes in gene interactions and changes in expression levels are needed. RESULTS: : In this article, we propose a differential network analysis method that considers the differential expression of individual genes when identifying differential edges. First, two hypothesis test statistics are used to quantify changes in partial correlations between gene pairs and changes in expression levels for individual genes. Then, an optimization framework is proposed to combine the two test statistics so that the resulting differential network has a hierarchical property, where a differential edge can be considered only if at least one of the two involved genes is differentially expressed. Simulation results indicate that our method outperforms current state-of-the-art methods. We apply our method to identify the differential networks between the luminal A and basal-like subtypes of breast cancer and those between acute myeloid leukemia and normal samples. Hub nodes in the differential networks estimated by our method, including both differentially and nondifferentially expressed genes, have important biological functions. AVAILABILITY AND IMPLEMENTATION: All the datasets underlying this article are publicly available. Processed data and source code can be accessed through the Github repository at https://github.com/Zhangxf-ccnu/chNet. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Reliable predictive models of protein-ligand binding affinity are required in many areas of biomedical research. Accurate prediction based on current descriptors or molecular fingerprints (FPs) remains a challenge. We develop novel interaction FPs (IFPs) to encode protein-ligand interactions and use them to improve the prediction. RESULTS: Proteo-chemometrics IFPs (PrtCmm IFPs) formed by combining extended connectivity fingerprints (ECFPs) with the proteo-chemometrics concept. Combining PrtCmm IFPs with machine-learning models led to efficient scoring models, which were validated on the PDBbind v2019 core set and CSAR-HiQ sets. The PrtCmm IFP Score outperformed several other models in predicting protein-ligand binding affinities. Besides, conventional ECFPs were simplified to generate new IFPs, which provided consistent but faster predictions. The relationship between the base atom properties of ECFPs and the accuracy of predictions was also investigated. AVAILABILITY: PrtCmm IFP has been implemented in the IFP Score Toolkit on github (https://github.com/debbydanwang/IFPscore). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

To reveal how gene regulatory networks change over cancer development, multiple time-varying differential networks between adjacent cancer stages should be estimated simultaneously. Since the network rewiring may be driven by the perturbation of certain individual genes, there may be some hub nodes shared by these differential networks. Although several methods have been developed to estimate differential networks from gene expression data, most of them are designed for estimating a single differential network, which neglect the similarities between different differential networks. In this article, we propose a new Gaussian graphical model-based method to jointly estimate multiple time-varying differential networks for identifying network rewiring over cancer development. A D-trace loss is used to determine the differential networks. A tree-structured group Lasso penalty is designed to identify the common hub nodes shared by different differential networks and the specific hub nodes unique to individual differential networks. Simulation experiment results demonstrate that our method outperforms other state-of-the-art techniques in most cases. We also apply our method to The Cancer Genome Atlas data to explore gene network rewiring over different breast cancer stages. Hub nodes in the estimated differential networks rediscover well known genes associated with the development and progression of breast cancer.

Most superpixel methods are sensitive to noise and cannot control the superpixel number precisely. To solve these problems, in this article, we propose a robust superpixel method called fuzzy simple linear iterative clustering (Fuzzy SLIC), which adopts a local spatial fuzzy C-means clustering and dynamic fuzzy superpixels. We develop a fast and precise superpixel number control algorithm called onion peeling (OP) algorithm. Fuzzy SLIC is insensitive to most types of noise, including Gaussian, salt and pepper, and multiplicative noise. The OP algorithm can control the superpixel number accurately without reducing much computational efficiency. In the validation experiments, we tested the Fuzzy SLIC and OP algorithm and compared them with state-of-the-art methods on the BSD500 and Pascal VOC2007 benchmarks. The experiment results show that our methods outperform state-of-the-art techniques in both noise-free and noisy environments.

Conditional maximum mean discrepancy (CMMD) can capture the discrepancy between conditional distributions by drawing support from nonlinear kernel functions; thus, it has been successfully used for pattern classification. However, CMMD does not work well on complex distributions, especially when the kernel function fails to correctly characterize the difference between intraclass similarity and interclass similarity. In this paper, a new kernel learning method is proposed to improve the discrimination performance of CMMD. It can be operated with deep network features iteratively and thus denoted as KLN for abbreviation. The CMMD loss and an autoencoder (AE) are used to learn an injective function. By considering the compound kernel, that is, the injective function with a characteristic kernel, the effectiveness of CMMD for data category description is enhanced. KLN can simultaneously learn a more expressive kernel and label prediction distribution; thus, it can be used to improve the classification performance in both supervised and semisupervised learning scenarios. In particular, the kernel-based similarities are iteratively learned on the deep network features, and the algorithm can be implemented in an end-to-end manner. Extensive experiments are conducted on four benchmark datasets, including MNIST, SVHN, CIFAR-10, and CIFAR-100. The results indicate that KLN achieves the state-of-the-art classification performance.

Establishing correspondence between two given geometrical graph structures is an important problem in computer vision and pattern recognition. In this paper, we propose a robust graph matching (RGM) model to improve the effectiveness and robustness on the matching graphs with deformations, rotations, outliers, and noise. First, we embed the joint geometric transformation into the graph matching model, which performs unary matching over graph nodes and local structure matching over graph edges simultaneously. Then, the L2,1-norm is used as the similarity metric in the presented RGM to enhance the robustness. Finally, we derive an objective function which can be solved by an effective optimization algorithm, and theoretically prove the convergence of the proposed algorithm. Extensive experiments on various graph matching tasks, such as outliers, rotations, and deformations show that the proposed RGM model achieves competitive performance compared to the existing methods.

Reconstructing gene networks from gene expression data is a long-standing challenge. In most applications, the observations can be divided into several distinct but related subpopulations and the gene expression measurements can be collected from multiple data types. Most existing methods are designed to estimate a single gene network from a single dataset. These methods may be suboptimal since they do not exploit the similarities and differences among different subpopulations and data types. In this article, we propose a joint graphical model to estimate the multiple gene networks simultaneously. Our model decomposes each subpopulation-specific gene network as a sum of common and unique components and imposes a group lasso penalty on gene networks corresponding to different data types. The gene network variations across subpopulations can be learned automatically by the decompositions of networks, and the similarities and differences among data types can be captured by the group lasso penalty. The simulation studies demonstrate that our method outperforms the state-of-the-art methods. We also apply our method to the cancer genome atlas breast cancer datasets to reconstruct subtype-specific gene networks. Hub nodes in the estimated subnetworks unique to individual cancer subtypes rediscover well-known genes associated with breast cancer subtypes and provide interesting predictions.

The procedure of establishing the correspondence between two sets of feature points is important in computer vision applications. In this article, an elastic net constraint-based tensor model is proposed for high-order graph matching. To control the tradeoff between the sparsity and the accuracy of the matching results, an elastic net constraint is introduced into the tensor-based graph matching model. Then, a nonmonotone spectral projected gradient (NSPG) method is derived to solve the proposed matching model. During the optimization of using NSPG, we propose an algorithm to calculate the projection on the feasible convex sets of elastic net constraint. Further, the global convergence of solving the proposed model using the NSPG method was proved. The superiority of the proposed method is verified through experiments on the synthetic data and natural images.

Non-small cell lung cancer (NSCLC) caused by mutation of the epidermal growth factor receptor (EGFR) is a major cause of death worldwide. Tyrosine kinase inhibitors (TKIs) of EGFR have been developed and show promising results at the initial stage of therapy. However, in most cases, their efficacy becomes limited due to the emergence of secondary mutations causing drug resistance after about a year. In this work, we investigated the mechanism of drug resistance due to these mutations. We performed molecular dynamics (MD) simulations of EGFR-drug interactions to obtain Euclidean distance and binding free energy values to analyse drug resistance and visualize drug-protein interactions. A PCA-based method is proposed to find normal, rigid, flexible, and critical residues. We have established a systematic method for the visualization of protein-drug interactions, which provides an effective framework for the analysis of drug resistance in lung cancer at the atomic level.

Most existing saliency methods measure fore- ground saliency by using the contrast of a foreground region to its local context, or boundary priors and spatial compactness. These methods are not powerful enough to extract a precise salient region from noisy and cluttered backgrounds. To evaluate the contrast of salient and background regions effectively, we consider high-level features from both supervised and unsupervised methods. We propose an affinity-based robust background subtraction technique and maximum attention map using a pre-trained convolution neural network. This affinity-based technique uses pixel similarities to propagate the values of salient pixels among foreground and background regions and their union. The salient pixel value controls the foreground and background information by using multiple pixel affinities. The maximum attention map is derived from the convolution neural network using features of the Pooling and Relu layers. This method can detect salient regions from images that have noisy and cluttered backgrounds. Our experimental results demonstrate the effectiveness of the proposed approach on six different saliency data sets and benchmarks and show that it improves the quality of detection beyond current saliency detection methods.

Nowadays, many data are multidimensional, which are called tensors. Tensor computations have been applied in different fields and various software libraries have been developed. However, not much attention has been received for developing a hardware architecture to accelerate the tensor computations. In this article, an efficient and unified processing element (PE) array for the 3-D tensor computation is demonstrated. Our PE array is optimized for thin and tall tensor-matrix multiplication and two types of tensor times matrices chain (TTMc) operations. Our design is evaluated in three study cases and compared with the state-of-the-art design. By using computation partition and rearrangement, data movement between the field-programmable gate array (FPGA) and off-chip DDR memory can be reduced by O(I2), where I is the maximum range among all the dimensions of the data tensor. For TTMc implementation, clock frequency has been increased by 18% compared with the state-of-the-art implementation on the same FPGA chip. An experiment on 3-D volumetric data set rendering by tensor approximation method is conducted for demonstration. For the bricks reconstruction process, the runtime decreased by 50%, i.e., two times faster, on our FPGA implementation compared with that running on GPU. In CANDECOMP/PARAFAC decomposition, for one iteration, the runtime has been decreased by up to 93% compared with the programs implemented by Tensorly, which is a python library.

Deep supervised hashing for image retrieval has attracted researchers' attention due to its high efficiency and superior retrieval performance. Most existing deep supervised hashing works, which are based on pairwise/triplet labels, suffer from the expensive computational cost and insufficient utilization of the semantics information. Recently, deep classwise hashing introduced a classwise loss supervised by class labels information alternatively; however, we find it still has its drawback. In this paper, we propose an improved deep classwise hashing, which enables hashing learning and class centers learning simultaneously. Specifically, we design a two-step strategy on center similarity learning. It interacts with the classwise loss to attract the class center to concentrate on the intra-class samples while pushing other class centers as far as possible. The centers similarity learning contributes to generating more compact and discriminative hashing codes. We conduct experiments on three benchmark datasets. It shows that the proposed method effectively surpasses the original method and outperforms state-of-the-art baselines under various commonly-used evaluation metrics for image retrieval.

MOTIVATION: Single-cell RNA sequencing (scRNA-seq) methods make it possible to reveal gene expression patterns at single-cell resolution. Due to technical defects, dropout events in scRNA-seq will add noise to the gene-cell expression matrix and hinder downstream analysis. Therefore, it is important for recovering the true gene expression levels before carrying out downstream analysis. RESULTS: In this article, we develop an imputation method, called scTSSR, to recover gene expression for scRNA-seq. Unlike most existing methods that impute dropout events by borrowing information across only genes or cells, scTSSR simultaneously leverages information from both similar genes and similar cells using a two-side sparse self-representation model. We demonstrate that scTSSR can effectively capture the Gini coefficients of genes and gene-to-gene correlations observed in single-molecule RNA fluorescence in situ hybridization (smRNA FISH). Down-sampling experiments indicate that scTSSR performs better than existing methods in recovering the true gene expression levels. We also show that scTSSR has a competitive performance in differential expression analysis, cell clustering and cell trajectory inference. AVAILABILITY AND IMPLEMENTATION: The R package is available at https://github.com/Zhangxf-ccnu/scTSSR. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Reconstruction of cancer gene networks from gene expression data is important for understanding the mechanisms underlying human cancer. Due to heterogeneity, the tumor tissue samples for a single cancer type can be divided into multiple distinct subtypes (inter-tumor heterogeneity) and are composed of non-cancerous and cancerous cells (intra-tumor heterogeneity). If tumor heterogeneity is ignored when inferring gene networks, the edges specific to individual cancer subtypes and cell types cannot be characterized. However, most existing network reconstruction methods do not simultaneously take inter-tumor and intra-tumor heterogeneity into account. RESULTS: In this article, we propose a new Gaussian graphical model-based method for jointly estimating multiple cancer gene networks by simultaneously capturing inter-tumor and intra-tumor heterogeneity. Given gene expression data of heterogeneous samples for different cancer subtypes, a non-cancerous network shared across different cancer subtypes and multiple subtype-specific cancerous networks are estimated jointly. Tumor heterogeneity can be revealed by the difference in the estimated networks. The performance of our method is first evaluated using simulated data, and the results indicate that our method outperforms other state-of-the-art methods. We also apply our method to The Cancer Genome Atlas breast cancer data to reconstruct non-cancerous and subtype-specific cancerous gene networks. Hub nodes in the networks estimated by our method perform important biological functions associated with breast cancer development and subtype classification. AVAILABILITY AND IMPLEMENTATION: The source code is available at https://github.com/Zhangxf-ccnu/NETI2. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

In this paper, we consider the multiclass clustering problem involving a hypergraph model. Fundamentally, we study a new normalized Laplacian tensor of an even-uniform weighted hypergraph. The hypergraph's connectivity is related with the second smallest Z-eigenvalue of the proposed Laplacian tensor. Particularly, an analogue of fractional Cheeger inequality holds. Next, we generalize the Laplacian tensor based approach from biclustering to multiclass clustering. A tensor optimization model with an orthogonal constraint is established and analyzed. Finally, we apply our hypergraph clustering approach to image segmentation and motion segmentation problems. Experimental results demonstrate that our method is effective.

Facial expressions are a strong visual intimation of gestural behaviors. The intelligent ability to learn these non-verbal cues of the humans is the key characteristic to develop efficient human computer interaction systems. Extracting an effective representation from facial expression images is a crucial step that impacts the recognition accuracy. In this paper, we propose a novel feature selection strategy using singular value decomposition (SVD) based co-clustering to search for the most salient regions in terms of facial features that possess a high discriminating ability among all expressions. To the best of our knowledge, this is the first known attempt to explicitly perform co-clustering in the facial expression recognition domain. In our method, Gabor filters are used to extract local features from an image and then discriminant features are selected based on the class membership in co-clusters. Experiments demonstrate that co-clustering localizes the salient regions of the face image. Not only does the procedure reduce the dimensionality but also improves the recognition accuracy. Experiments on CK plus, JAFFE and MMI databases validate the existence and effectiveness of these learned facial features.

The development and prognosis of complex diseases usually involves changes in regulatory relationships among biomolecules. Understanding how the regulatory relationships change with genetic alterations can help to reveal the underlying biological mechanisms for complex diseases. Although several models have been proposed to estimate the differential network between two different states, they are not suitable to deal with situations where the molecules of interest are affected by other covariates. Nor can they make use of prior information that provides insights about the structures of biomolecular networks. In this study, we introduce a novel weighted fused conditional Gaussian graphical model to jointly estimate two state-specific biomolecular regulatory networks and their difference between two different states. Unlike previous differential network estimation methods, our model can take into account the related covariates and the prior network information when inferring differential networks. The effectiveness of our proposed model is first evaluated based on simulation studies. Experiment results demonstrate that our model outperforms other state-of-the-art differential networks estimation models in all cases. We then apply our model to identify the differential gene network between two subtypes of glioblastoma based on gene expression and miRNA expression data. Our model is able to discover known mechanisms of glioblastoma and provide interesting predictions.

In the age of personalized medicine, there is a great need to classify cancer (from the same organ site) into homogeneous subtypes. Recent technology advancements in genome-wide molecular profiling have made it possible to profiling multiple molecular datasets to characterize the genomic changes in various cancer types. How to take full advantage of the availability of these omics data? And how to integrate these molecular data with patient clinical data to do a more systematic subtyping of cancer are the focuses of the paper. We proposed a new method called Molecular and Clinical Networks Fusion (MCNF) to classify cancer into homogeneous subtypes. Our method has two highlights: one is that it can integrate both numerical and non-numerical data into the fused network; the next highlight is that it is unsupervised, which means it can automatically determine the optimal number of clusters.

Establishing correspondence between pictorial descriptions of two images is an important task and can be treated as graph matching problem. However, the process of extracting a favourable graph structure from raw images for matching is influenced by cluttered backgrounds and deformations, which may result in the abundance of noisy graph structures. This paper addresses the problem of point set correspondence and presents a robust graph matching method which recovers the correspondence matches among the graph nodes in a CUR based factorization framework. The graph representation in terms of CUR, inherently preserves the actual nodes connection in sparse manner, this particularly renders the complex space-time realization of affinities among graph nodes. The reformulation of graph matching in terms of small CUR factorization matrices, allows to compute and relax the partially observed graphs, without observing the whole large-scale graph matrix. In particular, we propose two variants of this approach, first, approximating the matching matrix from small CUR observed graph structure, and second, completing the graph structure with higher order CUR form to find correspondence. The CUR based matching algorithms are realized by computing set of compatibility coefficients from pairwise matching graphs and further conducting the probability relaxation procedure to find the matching confidences among nodes. Experiments and analysis on synthetic and natural images dataset prove the effectiveness of proposed methods against state-of-the-art methods. We also explore CUR matching for non-rigid moving object in a video sequence to demonstrate the potential application of graph matching to video analysis.

Learning domain adaptive features aims to enhance the classification performance of the target domain by exploring the discriminant information from an auxiliary source set. Let X denote the feature and Y as the label. The most typical problem to be addressed is that PXYhas a so large variation between different domains that classification in the target domain is difficult. In this paper, we study the generalized conditional domain adaptation (DA) problem, in which both PYand PX|Ychange across domains, in a causal perspective. We propose transforming the class conditional probability matching to the marginal probability matching problem, under a proper assumption. We build an intermediate domain by employing a regression model. In order to enforce the most relevant data to reconstruct the intermediate representations, a low-rank constraint is placed on the regression model for regularization. The low-rank constraint underlines a global algebraic structure between different domains, and stresses the group compactness in representing the samples. The new model is considered under the discriminant subspace framework, which is favorable in simultaneously extracting the classification information from the source domain and adaptation information across domains. The model can be solved by an alternative optimization manner of quadratic programming and the alternative Lagrange multiplier method. To the best of our knowledge, this paper is the first to exploit low-rank representation, from the source domain to the intermediate domain, to learn the domain adaptive features. Comprehensive experimental results validate that the proposed method provides better classification accuracies with DA, compared with well-established baselines.

In this article, we propose a novel tensorial approach, namely, generalized tensor regression, for hyperspectral image classification. First, a simple and effective classifier, i.e., the ridge regression for multivariate labels, is extended to its tensorial version by taking advantages of tensorial representation. Then, the discrimination information of different modes is exploited to further strengthen the capacity of the model. Moreover, the model can be simplified and solved easily. Different from traditional tensorial methods, the proposed model can be utilized to capture not only the intrinsic structure of data in a physical sense but also the generalized relationship of data in a logical sense. Our proposed approach is shown to be effective for different classification purposes on a series of instantiations. Specifically, our experiment results with hyperspectral images collected by the airborne visible/infrared imaging spectrometer, the reflective optics spectrographic imaging system and the ITRES CASI-1500 demonstrate the effectiveness of the proposed approach as compared to other tensor-based classifiers and multiple kernel learning methods.

Hyperspectral image super-resolution addresses the problem of fusing a low-resolution hyperspectral image (LR-HSI) and a high-resolution multispectral image (HR-MSI) to produce a high-resolution hyperspectral image (HR-HSI). In this paper, we propose a novel fusion approach for hyperspectral image super-resolution by exploiting the specific properties of matrix decomposition, which consists of four main steps. First, an endmember extraction algorithm is used to extract an initial spectral matrix from LR-HSI. Then, with the initial spectral matrix, we estimate the spatial matrix, i.e., the spatial-contextual information, from the degraded observations of HR-HSI. Third, the spatial matrix is further utilized to estimate the spectral matrix from LR-HSI by solving a least squares (LS)-based problem. Finally, the target HR-HSI is constructed by combing the estimated spectral and spatial matrixes. In particular, two models are proposed to estimate the spatial matrix. One is a simple case that involves a LS-based problem, and the other is an elaborate case that consists of two fidelity terms and a spatial regularizer, where the spatial regularizer aiming to restrain the range of solutions is achieved by exploiting the superpixel-level low-rank characteristics of HR-HSI. Experiment results conducted on both synthetic and real data sets demonstrate the effectiveness of the proposed approach as compared to other hyperspectral image super-resolution methods.

Image set recognition has been widely applied in many practical problems like real-time video retrieval and image caption tasks. Due to its superior performance, it has grown into a significant topic in recent years. However, images with complicated variations, e.g., postures and human ages, are difficult to address, as these variations are continuous and gradual with respect to image appearance. Consequently, the crucial point of image set recognition is to mine the intrinsic connection or structural information from the image batches with variations. In this work, a Discriminant Residual Analysis (DRA) method is proposed to improve the classification performance by discovering discriminant features in related and unrelated groups. Specifically, DRA attempts to obtain a powerful projection which casts the residual representations into a discriminant subspace. Such a projection subspace is expected to magnify the useful information of the input space as much as possible, then the relation between the training set and the test set described by the given metric or distance will be more precise in the discriminant subspace. We also propose a nonfeasance strategy by defining another approach to construct the unrelated groups, which help to reduce furthermore the cost of sampling errors. Two regularization approaches are used to deal with the probable small sample size problem. Extensive experiments are conducted on benchmark databases, and the results show superiority and efficiency of the new methods.

In this article, the problem of automatically learning multiple parameters for kernel collaborative representation classification (KCRC) is considered. We investigate the KCRC and measure its generalization error via leave-one-out cross-validation (LOO-CV). By taking advantage of the specific properties of KCRC, a closed-form expression is derived for the outputs of LOO-CV. Then, a simple classification rule that provides probabilistic outputs is adopted, and thereby, an effective loss function that is an explicit function with respect to the parameters is proposed as the generalization error. The gradients of the loss function are calculated, and the parameters are learned by minimizing the loss function using a gradient-based optimization algorithm. Furthermore, the proposed approach makes it possible to solve the multiple kernel/feature learning problems of KCRC effectively. Experiment results on six data sets taken from different scenes demonstrate the effectiveness of the proposed approach.

Deep supervised hashing has emerged as an effective solution to large-scale semantic image retrieval problems in computer vision. Convolutional neural network-based hashing methods typically seek pairwise or triplet labels to conduct similarity-preserving learning. However, complex semantic concepts of visual contents are hard to capture by similar/dissimilar labels, which limits the retrieval performance. Generally, pairwise or triplet losses not only suffer from expensive training costs but also lack sufficient semantic information. In this paper, we propose a novel deep supervised hashing model to learn more compact class-level similarity-preserving binary codes. Our model is motivated by deep metric learning that directly takes semantic labels as supervised information in training and generates corresponding discriminant hashing code. Specifically, a novel cubic constraint loss function based on Gaussian distribution is proposed, which preserves semantic variations while penalizes the overlapping part of different classes in the embedding space. To address the discrete optimization problem introduced by binary codes, a two-step optimization strategy is proposed to provide efficient training and avoid the problem of gradient vanishing. Extensive experiments on five large-scale benchmark databases show that our model can achieve the state-of-the-art retrieval performance.

In the past decade, tensor computation is widely used in different areas. Various software toolbox have been released to assist tensor computation. However, there is still no hardware architecture to accelerate the tensor computation. This paper presents an efficient application specific instruction set processor (ASIP) for tensor computation. Different tensor computations are fully optimized in terms of resource usage and performance. We implement the ASIP on FPGA platform. We test our design by implementing the CANDECOMP/PARAFAC(CP) decomposition. Our design can achieve a low resource usage and run at 141 Mhz.

Lung cancer is the predominant reason for cancer deaths. Deletion mutation (del_E746-A750) in the Epidermal growth factor receptor (EGFR) is liable for 40% of Non-small cell lung cancer (NSCLC). However, 70% of the NSCLC active patients acquire T790M drug resistance mutation after progressing with the first-line EGFR Tyrosine kinase inhibitor (TKI). There are many third-generation EGFR-TKIs which are EGFR mutant selective and produced to heal the patients with T790M resistance mutation. Osimertinib is one of the third-generation EGFR TKIs which irreversibly inhibits the EGFR activity after the T790M mutation. Unfortunately, despite having an impressive initial response, 6 out of 15 patients who were diagnosed with third-generation EGFR-TKI would develop a new resistance and the most frequent being C797S mutation at exon 20. Numerous treatment techniques were implemented for patients who have progressed with C797S second resistance mutation, but there is no official fourth-generation EGFR TKI has invented. In our research work, we analyzed the stability changes for each structure in terms of reduction and hydrogen bonds. Our findings provide insight into the diversity of mechanisms through hydrogen bond analysis in different EGFR structures in terms of stability check and highlight the need for therapeutics and fourth-generation TKI to overcome resistance arbitrated by EGFR C797S.

Non-small cell lung cancer (NSCLC) with activating epidermal growth factor receptor (EGFR) is a major cause of death worldwide. Tyrosine kinase inhibitors (TKIs) have been developed to target the EGFR, stop the downstream signaling and the tumor growth. Despite of initial good results, drug resistance is developed after one year due to a secondary mutation. The L858R, T790M mutation and their combination change the conformational redistribution of EGFR. To combat drug resistance caused by the T790M mutation, AZD9291 third generation drug was approved by the food and drug administration agency, FDA, USA. However, resistance to AZD9291 is developed due to C797S mutation. In this paper, we investigate the drug resistance due to these genomic variations. We perform molecular dynamics (MD) simulation for EGFR, EGFR with L858R single point mutation, EGFR with L858R and T790M double point mutation and EGFR with L858R, T790M and C797S triple point mutation. We apply principal component analysis PCA and clustering to the atomic trajectories of EGFR and its mutants and extract the dominant motions. The first PC captures 29.04%, 51.17% 53.79% and 51.67% variance in WT, L858R, T790M and C797S mutants, respectively. First 20 PCs are used to explain the dynamics of the system, that captures about 90% of the variance in the system. This shows that the mutation increases the variance which leads to structural and dynamical changes and can be one of the reasons for the drug resistance. Our results provide new insights to the conformational dynamics and structural changes in EGFR and its mutants, that can be helpful for understanding the drug resistance mechanism and designing future therapies for NSCLC patients.

Superpixel-based fast fuzzy C-means clustering (SFFCM) is an efficient method for color image segmentation. However, it is sensitive to noise and blur. Its superpixel method called multiscale morphological gradient reconstruction (MMGR) is time consuming. In this paper, we propose an improved SFFCM method (ISFFCM) which replaces the MMGR in SFFCM with fuzzy simple linear iterative clustering (Fuzzy SLIC). Fuzzy SLIC is faster and more robust than MMGR for most types of noise, including salt and pepper noise, Gaussian noise and multiplicative noise. It is also more robust to image blur. In the validation experiments, we tested ISFFCM and SFFCM on the Berkeley benchmark. The experiment results show that our method outperforms SFFCM under noise and blurring environments.

Dynamic facial expression recognition has many useful applications in social networks, multimedia content analysis, security systems and others. This challenging process must be done under recurrent problems of image illumination and low resolution which changes at partial occlusions. This paper aims to produce a new facial expression recognition method based on the changes in the facial muscles. The geometric features are used to specify the facial regions i.e., mouth, eyes, and nose. The generic Fourier shape descriptor in conjunction with elliptic Fourier shape descriptor is used as an attribute to represent different emotions under frequency spectrum features. Afterwards a multi-class support vector machine is applied for classification of seven human expression. The statistical analysis showed our approach obtained overall competent recognition using 5-fold cross validation with high accuracy on well-known facial expression dataset.

In this paper, we investigate the ridge regression for multivariate labels by modelling each pixel and its surrounding pixels as a 3D tensor, and thereby propose a tensor ridge regression approach (TRR) for spatial-spectral hyperspectral image classification. Compared with the traditional ridge regression model, not only the spatial information is incorporated, but also the intrinsic spatial-spectral structure is captured. Moreover, the proposed TRR method is universal that it can be adopted to deal with the fusion of multiscale features for classification purpose. Experiment results conducted on two hyperspectral scenes demonstrate the effectiveness of the proposed method.

In this paper, we propose an effective training method to improve the performance of L2-normalized softmax for convolutional neural networks. Recent studies of deep learning show that by L2-normalizing the input features of softmax, the accuracy of CNN can be increased. Several works proposed novel loss functions based on the L2-normalized softmax. A common property shared by these modified normalized softmax models is that an extra set of parameters is introduced as the class centers. Although the physical meaning of this parameter is clear, few attentions have been paid to how to learn these class centers, which limits further improvement. In this paper, we address the problem of learning the class centers in the L2-normalized softmax. By treating the CNN training process as a time series, we propose a novel learning algorithm that combines the generally used gradient descent with the exponential moving average. Extensive experiments show that our model not only achieves better performance but also has a higher tolerance to the imbalance data.

Cancer is a collection of genetic diseases, with large phenotypic differences and genetic heterogeneity between different types of cancers and even within the same cancer type. Recent advances in genome-wide profiling provide an opportunity to investigate global molecular changes during the development and progression of cancer. Meanwhile, numerous statistical and machine learning algorithms have been designed for the processing and interpretation of high-throughput molecular data. Molecular subtyping studies have allowed the allocation of cancer into homogeneous groups that are considered to harbor similar molecular and clinical characteristics. Furthermore, this has helped researchers to identify both actionable targets for drug design as well as biomarkers for response prediction. In this review, we introduce five frequently applied techniques for generating molecular data, which are microarray, RNA sequencing, quantitative polymerase chain reaction, NanoString and tissue microarray. Commonly used molecular data for cancer subtyping and clinical applications are discussed. Next, we summarize a workflow for molecular subtyping of cancer, including data preprocessing, cluster analysis, supervised classification and subtype characterizations. Finally, we identify and describe four major challenges in the molecular subtyping of cancer that may preclude clinical implementation. We suggest that standardized methods should be established to help identify intrinsic subgroup signatures and build robust classifiers that pave the way toward stratified treatment of cancer patients.

SUMMARY: To identify biological network rewiring under different conditions, we develop a user-friendly R package, named DiffNetFDR, to implement two methods developed for testing the difference in different Gaussian graphical models. Compared to existing tools, our methods have the following features: (i) they are based on Gaussian graphical models which can capture the changes of conditional dependencies; (ii) they determine the tuning parameters in a data-driven manner; (iii) they take a multiple testing procedure to control the overall false discovery rate; and (iv) our approach defines the differential network based on partial correlation coefficients so that the spurious differential edges caused by the variants of conditional variances can be excluded. We also develop a Shiny application to provide easier analysis and visualization. Simulation studies are conducted to evaluate the performance of our methods. We also apply our methods to two real gene expression datasets. The effectiveness of our methods is validated by the biological significance of the identified differential networks. AVAILABILITY AND IMPLEMENTATION: R package and Shiny app are available at https://github.com/Zhangxf-ccnu/DiffNetFDR. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

SUMMARY: Imputation of dropout events that may mislead downstream analyses is a key step in analyzing single-cell RNA-sequencing (scRNA-seq) data. We develop EnImpute, an R package that introduces an ensemble learning method for imputing dropout events in scRNA-seq data. EnImpute combines the results obtained from multiple imputation methods to generate a more accurate result. A Shiny application is developed to provide easier implementation and visualization. Experiment results show that EnImpute outperforms the individual state-of-the-art methods in almost all situations. EnImpute is useful for correcting the noisy scRNA-seq data before performing downstream analysis. AVAILABILITY AND IMPLEMENTATION: The R package and Shiny application are available through Github at https://github.com/Zhangxf-ccnu/EnImpute. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Understanding the cellular architecture is a fundamental problem in various biological studies. C. elegans is widely used as a model organism in these studies because of its unique fate determinations. In recent years, researchers have worked extensively on C. elegans to excavate the regulations of genes and proteins on cell mobility and communication. Although various algorithms have been proposed to analyze nucleus, cell shape features are not yet well recorded. This paper proposes a method to systematically analyze three-dimensional morphological cellular features. RESULTS: Three-dimensional Membrane Morphological Segmentation (3DMMS) makes use of several novel techniques, such as statistical intensity normalization, and region filters, to pre-process the cell images. We then segment membrane stacks based on watershed algorithms. 3DMMS achieves high robustness and precision over different time points (development stages). It is compared with two state-of-the-art algorithms, RACE and BCOMS. Quantitative analysis shows 3DMMS performs best with the average Dice ratio of 97.7% at six time points. In addition, 3DMMS also provides time series of internal and external shape features of C. elegans. CONCLUSION: We have developed the 3DMMS based technique for embryonic shape reconstruction at the single-cell level. With cells accurately segmented, 3DMMS makes it possible to study cellular shapes and bridge morphological features and biological expression in embryo research.

Drug repositioning, i.e., identifying new indications for known drugs, has attracted a lot of attentions recently and is becoming an effective strategy in drug development. In literature, several computational approaches have been proposed to identify potential indications of old drugs based on various types of data sources. In this paper, by formulating the drug-disease associations as a low-rank matrix, we propose a novel method, namely DrPOCS, to identify candidate indications of old drugs based on projection onto convex sets (POCS). With the integration of drug structure and disease phenotype information, DrPOCS predicts potential associations between drugs and diseases with matrix completion. Benchmarking results demonstrate that our proposed approach outperforms popular existing approaches with high accuracy. In addition, a number of novel predicted indications are validated with various types of evidences, indicating the predictive power of our proposed approach.

The microRNAs (miRNAs), regulators of post-transcriptional processes, have been found to affect the efficacy of drugs by regulating the biological processes in which the target proteins of drugs may be involved. For example, some drugs develop resistance when certain miRNAs are overexpressed. Therefore, identifying miRNAs that affect drug effects can help understand the mechanisms of drug actions and design more efficient drugs. Although some computational approaches have been developed to predict miRNA-drug associations, such associations rarely provide explicit information about which miRNAs and how they affect drug efficacy. On the other hand, there are rich information about which miRNAs affect the efficacy of which drugs in the literature. In this paper, we present a novel text mining approach, named as EmDL (Extracting miRNA-Drug interactions from Literature), to extract the relationships of miRNAs affecting drug efficacy from literature. Benchmarking on the drug-miRNA interactions manually extracted from MEDLINE and PubMed Central, EmDL outperforms traditional text mining approaches as well as other popular methods for predicting drug-miRNA associations. Specifically, EmDL can effectively identify the sentences that describe the relationships of miRNAs affecting drug effects. The drug-miRNA interactome presented here can help understand how miRNAs affect drug effects and provide insights into the mechanisms of drug actions. In addition, with the information about drug-miRNA interactions, more effective drugs or combinatorial strategies can be designed in the future. The data used here can be accessed at http://mtd.comp-sysbio.org/.

Graphical models have been widely used to learn the conditional dependence structures among random variables. In many controlled experiments, such as the studies of disease or drug effectiveness, learning the structural changes of graphical models under two different conditions is of great importance. However, most existing graphical models are developed for estimating a single graph and based on a tacit assumption that there is no missing relevant variables, which wastes the common information provided by multiple heterogeneous data sets and underestimates the influence of latent/unobserved relevant variables. In this paper, we propose a joint differential network analysis (JDNA) model to jointly estimate multiple differential networks with latent variables from multiple data sets. The JDNA model is built on a penalized D-trace loss function, with group lasso or generalized fused lasso penalties. We implement a proximal gradient-based alternating direction method of multipliers to tackle the corresponding convex optimization problems. Extensive simulation experiments demonstrate that JDNA model outperforms state-of-the-art methods in estimating the structural changes of graphical models. Moreover, a series of experiments on several real-world data sets have been performed and experiment results consistently show that our proposed JDNA model is effective in identifying differential networks under different conditions.

Night-time vehicle detection is essential in building intelligent transportation systems (ITS) for road safety. Most of current night-time vehicle detection approaches focus on one or two classes of vehicles. In this paper, we present a novel multiclass vehicle detection system based on tensor decomposition and object proposal. Commonly used features such as histogram of oriented gradients and local binary pattern often produce useless image blocks (regions), which can result in unsatisfactory detection performance. Thus, we select blocks via feature ranking after tensor decomposition and only extract features from these selected blocks. To generate windows that contain all vehicles, we propose a novel object-proposal approach based on a state-of-the-art object-proposal method, local features, and image region similarity. The three terms are summed with learned weights to compute the reliability score of each proposal. A bio-inspired image enhancement method is used to enhance the brightness and contrast of input images. We have built a Hong Kong night-time multiclass vehicle dataset for evaluation. Our proposed vehicle detection approach can successfully detect four types of vehicles: 1) car; 2) taxi; 3) bus; and 4) minibus. Occluded vehicles and vehicles in the rain can also be detected. Our proposed method obtains 95.82% detection rate at 0.05 false positives per image, and it outperforms several state-of-the-art night-time vehicle detection approaches.

Extraction of salient object from blurred and similar background color image is very difficult task. Many image segmentation methods have been proposed to overcome this problem but their performance is unsatisfactory when the target object and background has similar color appearance. In this paper, we have proposed a technique to overcome this problem with fast fuzzy-c-mean membership maps. These maps are blended by using Porter-Duff compositing method. The composite process is accomplished under different blending modes where foreground element of one map blend on the dropback element of the second map. These blended maps contain some outliers, which are removed by applying morphological technique. Finally an image mask, which is the composite form of frequency prior, color prior and location prior of an image is used to extract the final salient map from the given blended maps. Experiments on four well-known datasets (MSRA, MSRA-1000, THUR15000 and SED) are conducted; The results indicate the efficiency of proposed method. Our approach produces more accurate image segmentation, where the background and foreground maps have similarity in color appearance.

Microscopic imaging has many applications in biological experiments. It is important to determine the cell locations in an image before other analysis tasks take place. In this paper, we propose a novel method to obtain cell and membrane segmentation based on the combination of k nearest neighbor clustering and biological morphology constraints. First, we produce preliminary segmentation with the watershed transformation. Then the segmentation results are optimized based on the morphological characteristics of the membrane. This method provides us with well-segmented cells and membranes, which significantly reduces errors in cell image analysis.

Summary: We develop DiffGraph, an R package that integrates four influential differential graphical models for identifying gene network rewiring under two different conditions from gene expression data. The input and output of different models are packaged in the same format, making it convenient for users to compare different models using a wide range of datasets and carry out follow-up analysis. Furthermore, the inferred differential networks can be visualized both non-interactively and interactively. The package is useful for identifying gene network rewiring from input datasets, comparing the predictions of different methods and visualizing the results. Availability and implementation: The package is available at https://github.com/Zhangxf-ccnu/DiffGraph. Contact: [email protected]. Supplementary information: Supplementary data are available at Bioinformatics online.

BACKGROUND: Non-small cell lung cancer (NSCLC) with activating EGFR mutations, especially exon 19 deletions and the L858R point mutation, is particularly responsive to gefitinib and erlotinib. However, the sensitivity varies for less common and rare EGFR mutations. There are various explanations for the low sensitivity of EGFR exon 20 insertions and the exon 20 T790 M point mutation to gefitinib/erlotinib. However, few studies discuss, from a structural perspective, why less common mutations, like G719X and L861Q, have moderate sensitivity to gefitinib/erlotinib. RESULTS: To decode the drug sensitivity/selectivity of EGFR mutants, it is important to analyze the interaction between EGFR mutants and EGFR inhibitors. In this paper, the 30 most common EGFR mutants were selected and the technique of protein-ligand interaction fingerprint (IFP) was applied to analyze and compare the binding modes of EGFR mutant-gefitinib/erlotinib complexes. Molecular dynamics simulations were employed to obtain the dynamic trajectory and a matrix of IFPs for each EGFR mutant-inhibitor complex. Multilinear Principal Component Analysis (MPCA) was applied for dimensionality reduction and feature selection. The selected features were further analyzed for use as a drug sensitivity predictor. The results showed that the accuracy of prediction of drug sensitivity was very high for both gefitinib and erlotinib. Targeted Projection Pursuit (TPP) was used to show that the data points can be easily separated based on their sensitivities to gefetinib/erlotinib. CONCLUSIONS: We can conclude that the IFP features of EGFR mutant-TKI complexes and the MPCA-based tensor object feature extraction are useful to predict the drug sensitivity of EGFR mutants. The findings provide new insights for studying and predicting drug resistance/sensitivity of EGFR mutations in NSCLC and can be beneficial to the design of future targeted therapies and innovative drug discovery.

Object proposal is one of the most key pre-processing steps for nighttime vehicle detection systems in intelligent transportation systems. However, most current object proposal methods are developed on daytime data sets, and these methods demonstrate unsatisfactory results when they are used on nighttime images. Therefore, this paper presents a novel Bayes saliency-based object proposal generator for nighttime RGB traffic images to generate a modest and accurate set of proposals, which are more likely to be vehicles for preceding vehicle detection. First, we propose a new Bayes saliency detection approach in which prior estimation, feature extraction, weight estimation, and Bayes rule are used to compute saliency maps. Then, we propose a simple but effective object proposal generator based on the Bayes saliency map. Multi-scale sliding window, proposal rejecting, scoring, and non-maximum suppression are combined to generate a modest and effective set of proposals. Experimental results demonstrate that our proposed approach generates a modest set of proposals and outperforms some state-of-the-art methods on nighttime images in terms of various evaluation metrics. Furthermore, our proposed object proposal approach can improve the detection performance and the speed of several state-of-the-art vehicle detection approaches.

MOTIVATION: Cell fate specification plays a key role to generate distinct cell types during metazoan development. However, most of the underlying signaling networks at cellular level are not well understood. Availability of time lapse single-cell gene expression data collected throughout Caenorhabditis elegans embryogenesis provides an excellent opportunity for investigating signaling networks underlying cell fate specification at systems, cellular and molecular levels. RESULTS: We propose a framework to infer signaling networks at cellular level by exploring the single-cell gene expression data. Through analyzing the expression data of nhr-25 , a hypodermis-specific transcription factor, in every cells of both wild-type and mutant C.elegans embryos through RNAi against 55 genes, we have inferred a total of 23 genes that regulate (activate or inhibit) nhr-25 expression in cell-specific fashion. We also infer the signaling pathways consisting of each of these genes and nhr-25 based on a probabilistic graphical model for the selected five founder cells, 'ABarp', 'ABpla', 'ABpra', 'Caa' and 'Cpa', which express nhr-25 and mostly develop into hypodermis. By integrating the inferred pathways, we reconstruct five signaling networks with one each for the five founder cells. Using RNAi gene knockdown as a validation method, the inferred networks are able to predict the effects of the knockdown genes. These signaling networks in the five founder cells are likely to ensure faithful hypodermis cell fate specification in C.elegans at cellular level. AVAILABILITY AND IMPLEMENTATION: All source codes and data are available at the github repository https://github.com/xthuang226/Worm_Single_Cell_Data_and_Codes.git . CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Understanding how gene regulatory networks change under different cellular states is important for revealing insights into network dynamics. Gaussian graphical models, which assume that the data follow a joint normal distribution, have been used recently to infer differential networks. However, the distributions of the omics data are non-normal in general. Furthermore, although much biological knowledge (or prior information) has been accumulated, most existing methods ignore the valuable prior information. Therefore, new statistical methods are needed to relax the normality assumption and make full use of prior information. RESULTS: We propose a new differential network analysis method to address the above challenges. Instead of using Gaussian graphical models, we employ a non-paranormal graphical model that can relax the normality assumption. We develop a principled model to take into account the following prior information: (i) a differential edge less likely exists between two genes that do not participate together in the same pathway; (ii) changes in the networks are driven by certain regulator genes that are perturbed across different cellular states and (iii) the differential networks estimated from multi-view gene expression data likely share common structures. Simulation studies demonstrate that our method outperforms other graphical model-based algorithms. We apply our method to identify the differential networks between platinum-sensitive and platinum-resistant ovarian tumors, and the differential networks between the proneural and mesenchymal subtypes of glioblastoma. Hub nodes in the estimated differential networks rediscover known cancer-related regulator genes and contain interesting predictions. AVAILABILITY AND IMPLEMENTATION: The source code is at https://github.com/Zhangxf-ccnu/pDNA. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: The accurate identification of protein complexes is important for the understanding of cellular organization. Up to now, computational methods for protein complex detection are mostly focus on mining clusters from protein-protein interaction (PPI) networks. However, PPI data collected by high-throughput experimental techniques are known to be quite noisy. It is hard to achieve reliable prediction results by simply applying computational methods on PPI data. Behind protein interactions, there are protein domains that interact with each other. Therefore, based on domain-protein associations, the joint analysis of PPIs and domain-domain interactions (DDI) has the potential to obtain better performance in protein complex detection. As traditional computational methods are designed to detect protein complexes from a single PPI network, it is necessary to design a new algorithm that could effectively utilize the information inherent in multiple heterogeneous networks. RESULTS: In this paper, we introduce a novel multi-network clustering algorithm to detect protein complexes from multiple heterogeneous networks. Unlike existing protein complex identification algorithms that focus on the analysis of a single PPI network, our model can jointly exploit the information inherent in PPI and DDI data to achieve more reliable prediction results. Extensive experiment results on real-world data sets demonstrate that our method can predict protein complexes more accurately than other state-of-the-art protein complex identification algorithms. CONCLUSIONS: In this work, we demonstrate that the joint analysis of PPI network and DDI network can help to improve the accuracy of protein complex detection.

We explore the drug resistance mechanism in non-small cell lung cancer treatment by characterizing the drug-binding site of a protein mutant based on local surface and energy features. These features are transformed to an eigen-binding site space and used for drug resistance level prediction and analysis.

Dimension reduction is an important topic in pattern analysis and machine learning, and it has wide applications in feature representation and pattern classification. In the past two decades, sliced inverse regression (SIR) has attracted much research efforts due to its effectiveness and efficacy in dimension reduction. However, two drawbacks limit further applications of SIR. First, the computation complexity of SIR is usually high in the situation of high-dimensional data. Second, sparsity of projection subspace is not well mined for improving the feature selection and model interpretation abilities. This paper proposes to compute the SIR projection vectors in the spectral space, then an approximated regression solution can be obtained with a faster speed. Moreover, the adaptive lasso is used to attain a sparse and globally optimal solution, which is important in variable selection. To complete the robust pattern classification task with corruptions, a correntropy-based and class-wise regression model is designed in this paper. It takes a smooth penalty instead of sparsity constraint in the regression coefficients, and it can be conducted in class-wise, thus it is more flexible in practice. Extensive experiments are conducted by using some real and benchmark data sets, e.g., high-dimensional facial images and gene microarray data, to evaluate the new algorithms. The new proposals attain competitive results and are compared with other state-of-the-art methods.

This paper presents an effective nighttime vehicle detection system that combines a novel bioinspired image enhancement approach with a weighted feature fusion technique. Inspired by the retinal mechanism in natural visual processing, we develop a nighttime image enhancement method by modeling the adaptive feedback from horizontal cells and the center-surround antagonistic receptive fields of bipolar cells. Furthermore, we extract features based on the convolutional neural network, histogram of oriented gradient, and local binary pattern to train the classifiers with support vector machine. These features are fused by combining the score vectors of each feature with the learnt weights. During detection, we generate accurate regions of interest by combining vehicle taillight detection with object proposals. Experimental results demonstrate that the proposed bioinspired image enhancement method contributes well to vehicle detection. Our vehicle detection method demonstrates a 95.95% detection rate at 0.0575 false positives per image and outperforms some state-of-the-art techniques. Our proposed method can deal with various scenes including vehicles of different types and sizes and those with occlusions and in blurred zones. It can also detect vehicles at various locations and multiple vehicles.

The detection of protein complexes from protein-protein interaction (PPI) networks is an important step toward understanding the functional organization within cells. A great number of graph clustering algorithms have been proposed to undertake this task. Since PPI data collected by high-throughput technologies is quite noisy, simply applying graph clustering algorithms on PPI data is generally not adequate to achieve reliable prediction results. Behind protein interactions, there are protein domains that interact with each other. Jointly exploiting protein-protein interactions and domain-domain interactions (DDI) have the potential to increase the accuracy of protein complex detection. However, traditional graph clustering algorithms focus on clustering proteins within a single PPI network, and cannot make use of information inherent in other heterogeneous networks. In this paper, we proposed a novel generative model to perform multi-network clustering. Unlike previous protein complex detection algorithms that can only utilize the information within a single PPI network, our model is a flexible framework that can take into account PPIs, DDIs and domain-protein associations to achieve more consistent and reliable clustering results. Experiment results on real data demonstrate that our method performs much better than state-of-the-art protein complex detection techniques.

BACKGROUND: Protein complexes carry out nearly all signaling and functional processes within cells. The study of protein complexes is an effective strategy to analyze cellular functions and biological processes. With the increasing availability of proteomics data, various computational methods have recently been developed to predict protein complexes. However, different computational methods are based on their own assumptions and designed to work on different data sources, and various biological screening methods have their unique experiment conditions, and are often different in scale and noise level. Therefore, a single computational method on a specific data source is generally not able to generate comprehensive and reliable prediction results. RESULTS: In this paper, we develop a novel Two-layer INtegrative Complex Detection (TINCD) model to detect protein complexes, leveraging the information from both clustering results and raw data sources. In particular, we first integrate various clustering results to construct consensus matrices for proteins to measure their overall co-complex propensity. Second, we combine these consensus matrices with the co-complex score matrix derived from Tandem Affinity Purification/Mass Spectrometry (TAP) data and obtain an integrated co-complex similarity network via an unsupervised metric fusion method. Finally, a novel graph regularized doubly stochastic matrix decomposition model is proposed to detect overlapping protein complexes from the integrated similarity network. CONCLUSIONS: Extensive experimental results demonstrate that TINCD performs much better than 21 state-of-the-art complex detection techniques, including ensemble clustering and data integration techniques.

BACKGROUND: Protein complexes are the key molecular entities to perform many essential biological functions. In recent years, high-throughput experimental techniques have generated a large amount of protein interaction data. As a consequence, computational analysis of such data for protein complex detection has received increased attention in the literature. However, most existing works focus on predicting protein complexes from a single type of data, either physical interaction data or co-complex interaction data. These two types of data provide compatible and complementary information, so it is necessary to integrate them to discover the underlying structures and obtain better performance in complex detection. RESULTS: In this study, we propose a novel multi-view clustering algorithm, called the Partially Shared Multi-View Clustering model (PSMVC), to carry out such an integrated analysis. Unlike traditional multi-view learning algorithms that focus on mining either consistent or complementary information embedded in the multi-view data, PSMVC can jointly explore the shared and specific information inherent in different views. In our experiments, we compare the complexes detected by PSMVC from single data source with those detected from multiple data sources. We observe that jointly analyzing multi-view data benefits the detection of protein complexes. Furthermore, extensive experiment results demonstrate that PSMVC performs much better than 16 state-of-the-art complex detection techniques, including ensemble clustering and data integration techniques. CONCLUSIONS: In this work, we demonstrate that when integrating multiple data sources, using partially shared multi-view clustering model can help to identify protein complexes which are not readily identifiable by conventional single-view-based methods and other integrative analysis methods. All the results and source codes are available on https://github.com/Oyl-CityU/PSMVC .

BACKGROUND: To facilitate advances in personalized medicine, it is important to detect predictive, stable and interpretable biomarkers related with different clinical characteristics. These clinical characteristics may be heterogeneous with respect to underlying interactions between genes. Usually, traditional methods just focus on detection of differentially expressed genes without taking the interactions between genes into account. Moreover, due to the typical low reproducibility of the selected biomarkers, it is difficult to give a clear biological interpretation for a specific disease. Therefore, it is necessary to design a robust biomarker identification method that can predict disease-associated interactions with high reproducibility. RESULTS: In this article, we propose a regularized logistic regression model. Different from previous methods which focus on individual genes or modules, our model takes gene pairs, which are connected in a protein-protein interaction network, into account. A line graph is constructed to represent the adjacencies between pairwise interactions. Based on this line graph, we incorporate the degree information in the model via an adaptive elastic net, which makes our model less dependent on the expression data. Experimental results on six publicly available breast cancer datasets show that our method can not only achieve competitive performance in classification, but also retain great stability in variable selection. Therefore, our model is able to identify the diagnostic and prognostic biomarkers in a more robust way. Moreover, most of the biomarkers discovered by our model have been verified in biochemical or biomedical researches. CONCLUSIONS: The proposed method shows promise in the diagnosis of disease pathogenesis with different clinical characteristics. These advances lead to more accurate and stable biomarker discovery, which can monitor the functional changes that are perturbed by diseases. Based on these predictions, researchers may be able to provide suggestions for new therapeutic approaches.

BACKGROUND: Several recent studies have used the Minimum Dominating Set (MDS) model to identify driver nodes, which provide the control of the underlying networks, in protein interaction networks. There may exist multiple MDS configurations in a given network, thus it is difficult to determine which one represents the real set of driver nodes. Because these previous studies only focus on static networks and ignore the contextual information on particular tissues, their findings could be insufficient or even be misleading. RESULTS: In this study, we develop a Collective-Influence-corrected Minimum Dominating Set (CI-MDS) model which takes into account the collective influence of proteins. By integrating molecular expression profiles and static protein interactions, 16 tissue-specific networks are established as well. We then apply the CI-MDS model to each tissue-specific network to detect MDS proteins. It generates almost the same MDSs when it is solved using different optimization algorithms. In addition, we classify MDS proteins into Tissue-Specific MDS (TS-MDS) proteins and HouseKeeping MDS (HK-MDS) proteins based on the number of tissues in which they are expressed and identified as MDS proteins. Notably, we find that TS-MDS proteins and HK-MDS proteins have significantly different topological and functional properties. HK-MDS proteins are more central in protein interaction networks, associated with more functions, evolving more slowly and subjected to a greater number of post-translational modifications than TS-MDS proteins. Unlike TS-MDS proteins, HK-MDS proteins significantly correspond to essential genes, ageing genes, virus-targeted proteins, transcription factors and protein kinases. Moreover, we find that besides HK-MDS proteins, many TS-MDS proteins are also linked to disease related genes, suggesting the tissue specificity of human diseases. Furthermore, functional enrichment analysis reveals that HK-MDS proteins carry out universally necessary biological processes and TS-MDS proteins usually involve in tissue-dependent functions. CONCLUSIONS: Our study uncovers key features of TS-MDS proteins and HK-MDS proteins, and is a step forward towards a better understanding of the controllability of human interactomes.

A new approach for text-independent phoneme segmentation at sampling point level is proposed in this paper. The algorithm consists of two phases: First, the voiced sections in speech data are detected using the information of vocal folds vibration contained in electroglottograph (EGG). A Hilbert envelope feature is adopted to achieve sampling point level detection accuracy. Second, the voiced sections and other sections are treated separately. Each voiced section is divided into several candidate phonemes using the Viterbi algorithm. Then adjacent candidate phonemes are merged based on a Hotellings T-square test method. For other sections, the unvoiced consonants are detected from silence based on a singularity exponent feature. Comparison experiments show that the proposed method has better performance than the existing ones for a variety of tolerances, and is more robust to noise.

Automatic cell tracking for time-lapse images becomes more and more important for live cell studies because the manual tracking is extremely time consuming. In this paper, we proposed a method for C.elegans cell tracking based on probabilistic relaxation labeling (PRL). The experiment results obtained in this research indicate that our method could track the C.elegans cells with a high accuracy at a very low time resolution. Our method provides an efficient tool for the analysis of high-throughput large C.elegans microscopy image data sets.

Recently, machine learning techniques have become an indispensable alternative for computational studies of cancers and efficient prediction of cancer-drug responses or drug resistance levels. Meanwhile, in cancer characterization, molecular dynamics (MD) simulations can greatly reveal the dynamic and functional features of cancer-related proteins. In our work, MD simulations were implemented to extract the EGFR TK mutation (dynamic) features of a non-small-cell lung cancer (NSCLC)-patient group. Specifically, the relative positions of a drug-binding site and a drug molecule in the dynamics-trajectory were calculated and used for characterizing the dynamic features. These derived features, couples with patient personal features, were subsequently handled by a model called SFABSRM, which combines Supervised Factor Analysis and Softmax Regression Model. SFABSRM first uses factor analysis to evaluate the contributions of the selected features, and in our analysis it suggested that dynamic features play an important role in correlating with the cancer-drug responses. Further, SFABSRM applies the regression model for a drug response prediction, which further verified the important contribution of dynamic characteristics to this prediction. The support vector machine (SVM) model was conducted as a comparison with SFABSRM, leading to an agreement with the earlier conclusion. Overall, these studies can greatly benefit the NSCLC studies and drug discovery.

This paper proposes local topology preserved features in graph matching problem based on tensor technique. Many tensor based works paid much attention on catching many invariant feature tuples while local information for every single point to improve matching performance is also important. Here our proposed Local Topology Preserved Tensor (LTPT) models not only take into account of the neighbor structure but also employ the three-order tensor technique to keep the geometric consistency. Extensive experiments on the synthetic and real datasets show that LTPT performs better than the state-of-the-art graph matching methods.

BACKGROUND: Epidermal growth factor receptor (EGFR) mutation-induced drug resistance has caused great difficulties in the treatment of non-small-cell lung cancer (NSCLC). However, structural information is available for just a few EGFR mutants. In this study, we created an EGFR Mutant Structural Database (freely available at http://bcc.ee.cityu.edu.hk/data/EGFR.html ), including the 3D EGFR mutant structures and their corresponding binding free energies with two commonly used inhibitors (gefitinib and erlotinib). RESULTS: We collected the information of 942 NSCLC patients belonging to 112 mutation types. These mutation types are divided into five groups (insertion, deletion, duplication, modification and substitution), and substitution accounts for 61.61% of the mutation types and 54.14% of all the patients. Among all the 942 patients, 388 cases experienced a mutation at residue site 858 with leucine replaced by arginine (L858R), making it the most common mutation type. Moreover, 36 (32.14%) mutation types occur at exon 19, and 419 (44.48%) patients carried a mutation at exon 21. In this study, we predicted the EGFR mutant structures using Rosetta with the collected mutation types. In addition, Amber was employed to refine the structures followed by calculating the binding free energies of mutant-drug complexes. CONCLUSIONS: The EGFR Mutant Structural Database provides resources of 3D structures and the binding affinity with inhibitors, which can be used by other researchers to study NSCLC further and by medical doctors as reference for NSCLC treatment.

As the data acquirement technologies develop rapidly, both the amount and types of data become larger and larger. However, noise and outliers usually attach to the data and then affect the real performance of leaning algorithms in data mining and pattern analysis. To address this problem, the importance of the sample itself in building the optimal subspace is explored, and then an importance-sampling-inspired method is proposed for outlier suppressing feature extraction. First, we assign each sample a weight, which is estimated by graph Laplacian, and then calculate the approximated mean for each subject. By highlighting the most subject-oriented samples, the weighted average and the scatter metrics can be measured with maximum margins and superior classification performance. The supervised information integrates local data structure with respective contributions to building the optimal subspace. The linear criterion can be extended to a nonlinear case by the kernel trick. A regularization framework is proposed to deal with the rank-deficient problem, which is usually induced by the small sample size of training set. Competitive performance of our algorithm has been validated by extensive experiments performed on the synthetic and benchmark data, including facial images and gene micro-array data.

In recent years, more 3D protein structures have become available, which has made the analysis of large molecular structures much easier. There is a strong demand for geometric models for the study of protein-related interactions. Alpha shape and Delaunay triangulation are powerful tools to represent protein structures and have advantages in characterizing the surface curvature and atom contacts. This review presents state-of-the-art applications of alpha shape and Delaunay triangulation in the studies on protein-DNA, protein-protein, protein-ligand interactions and protein structure analysis.

BACKGROUND: Identification of protein complexes can help us get a better understanding of cellular mechanism. With the increasing availability of large-scale protein-protein interaction (PPI) data, numerous computational approaches have been proposed to detect complexes from the PPI networks. However, most of the current approaches do not consider overlaps among complexes or functional annotation information of individual proteins. Therefore, they might not be able to reflect the biological reality faithfully or make full use of the available domain-specific knowledge. RESULTS: In this paper, we develop a Generative Model with Functional and Topological Properties (GMFTP) to describe the generative processes of the PPI network and the functional profile. The model provides a working mechanism for capturing the interaction structures and the functional patterns of proteins. By combining the functional and topological properties, we formulate the problem of identifying protein complexes as that of detecting a group of proteins which frequently interact with each other in the PPI network and have similar annotation patterns in the functional profile. Using the idea of link communities, our method naturally deals with overlaps among complexes. The benefits brought by the functional properties are demonstrated by real data analysis. The results evaluated using four criteria with respect to two gold standards show that GMFTP has a competitive performance over the state-of-the-art approaches. The effectiveness of detecting overlapping complexes is also demonstrated by analyzing the topological and functional features of multi- and mono-group proteins. CONCLUSIONS: Based on the results obtained in this study, GMFTP presents to be a powerful approach for the identification of overlapping protein complexes using both the PPI network and the functional profile. The software can be downloaded from http://mail.sysu.edu.cn/home/[email protected]/dai/others/GMFTP.zip.

Data generated from microarray experiments often suffer from missing values. As most downstream analyses need full matrices as input, these missing values have to be estimated. Bayesian principal component analysis (BPCA) is a well-known microarray missing value estimation method, but its performance is not satisfactory on datasets with strong local similarity structure. A bicluster-based BPCA (bi-BPCA) method is proposed in this paper to fully exploit local structure of the matrix. In a bicluster, the most correlated genes and experimental conditions with the missing entry are identified, and BPCA is conducted on these biclusters to estimate the missing values. An automatic parameter learning scheme is also developed to obtain optimal parameters. Experimental results on four real microarray matrices indicate that bi-BPCA obtains the lowest normalized root-mean-square error on 82.14% of all missing rates.

Unobtrusive, contactless recordings of physiological signals are very important for many health and human-computer interaction applications. Most current systems require sensors which intrusively touch the user's skin. Recent advances in contact-free physiological signals open the door to many new types of applications. This technology promises to measure heart rate (HR) and respiration using video only. The effectiveness of this technology, its limitations, and ways of overcoming them deserves particular attention. In this paper, we evaluate this technique for measuring HR in a controlled situation, in a naturalistic computer interaction session, and in an exercise situation. For comparison, HR was measured simultaneously using an electrocardiography device during all sessions. The results replicated the published results in controlled situations, but show that they cannot yet be considered as a valid measure of HR in naturalistic human-computer interaction. We propose a machine learning approach to improve the accuracy of HR detection in naturalistic measurements. The results demonstrate that the root mean squared error is reduced from 43.76 to 3.64 beats/min using the proposed method.

Biclustering is an important technique in data mining for searching similar patterns. Geometric biclustering (GBC) method is used to reduce the complexity of the NP-complete biclustering algorithm. This paper studies three commonly used modern platforms including multi-core CPU, GPU and FPGA to accelerate this GBC algorithm. By analyzing the parallelizing property of the GBC algorithm, we design 1) a multi-threaded software running on a server grade multi-core CPU system, 2) a CUDA program for GPU to accelerate the GBC algorithm, and 3) a novel parameterizable and scalable hardware architecture implemented on an FPGA. Genes microarray pattern analysis is employed as an example to demonstrate performance comparisons on different platforms. In particular, we compare the speed and energy efficiency of the three proposed methods. We found that 1) GPU achieves the highest average speedup of 48 × compared to single-threaded GBC program, 2) Our FPGA design can achieve higher speedup of 4 × for the computation for large microarray, and 3) FPGA consumes the least energy, which is about 3.53 × more efficient than the single-threaded GBC program.

Mutations in EGFR kinase domain can cause non-small-cell lung cancer, which is one of the most lethal diseases in the world. However, current therapy is limited by the drug resistance effect in different EGFR mutants. There is an urgent demand for developing computational methods to predict drug resisted mutations. In this study, we use quantum mechanics and molecular mechanics models to generate EGFR mutants, and apply molecular dynamic to simulate EGFR-drug interactions. Hydrogen bonds and binding free energy are used to reveal the underlying principle of drug resistance in EGFR. The results show that drug resisted mutants do not establish hydrogen bond between the drug and the protein molecule while having large binding free energy. These properties can be used to predict resistance to anti-EGFR drugs due to protein mutations.

BACKGROUND: Recently a series of algorithms have been developed, providing automatic tools for tracing C. elegans embryonic cell lineage. In these algorithms, 3D images collected from a confocal laser scanning microscope were processed, the output of which is cell lineage with cell division history and cell positions with time. However, current image segmentation algorithms suffer from high error rate especially after 350-cell stage because of low signal-noise ratio as well as low resolution along the Z axis (0.5-1 microns). As a result, correction of the errors becomes a huge burden. These errors are mainly produced in the segmentation of nuclei. Thus development of a more accurate image segmentation algorithm will alleviate the hurdle for automated analysis of cell lineage. RESULTS: This paper presents a new type of nuclei segmentation method embracing an bi-directional prediction procedure, which can greatly reduce the number of false negative errors, the most common errors in the previous segmentation. In this method, we first use a 2D region growing technique together with the level-set method to generate accurate 2D slices. Then a modified gradient method instead of the existing 3D local maximum method is adopted to detect all the 2D slices located in the nuclei center, each of which corresponds to one nucleus. Finally, the bi-directional prediction method based on the images before and after the current time point is introduced into the system to predict the nuclei in low quality parts of the images. The result of our method shows a notable improvement in the accuracy rate. For each nucleus, its precise location, volume and gene expression value (gray value) is also obtained, all of which will be useful in further downstream analyses. CONCLUSIONS: The result of this research demonstrates the advantages of the bi-directional prediction method in the nuclei segmentation over that of StarryNite/MatLab StarryNite. Several other modifications adopted in our nuclei segmentation system are also discussed.

Interactions between biomolecules play an essential role in various biological processes. For predicting DNA-binding or protein-binding proteins, many machine-learning-based techniques have used various types of features to represent the interface of the complexes, but they only deal with the properties of a single atom in the interface and do not take into account the information of neighborhood atoms directly. This paper proposes a new feature representation method for biomolecular interfaces based on the theory of graph wavelet. The enhanced graph wavelet features (EGWF) provides an effective way to characterize interface feature through adding physicochemical features and exploiting a graph wavelet formulation. Particularly, graph wavelet condenses the information around the center atom, and thus enhances the discrimination of features of biomolecule binding proteins in the feature space. Experiment results show that EGWF performs effectively for predicting DNA-binding and protein-binding proteins in terms of Matthew's correlation coefficient (MCC) score and the area value under the receiver operating characteristic curve (AUC).

Traditional signal processing methods cannot detect interspersed repeats and generally cannot handle nonstationary signals. In this paper, we propose a new method for periodicity detection in protein sequences to locate interspersed repeats. We first apply the autoregressive model with a sliding window to find possible repeating subsequences within a protein sequence. Then, we utilize an iterative hidden Markov model (HMM) to count the number of subsequences similar to each of the possible repeating subsequences. An iterative HMM search of the potential repeating subsequences can help identify interspersed repeats. Finally, the numbers of repeating subsequences are aggregated together as a feature and used in the classification process. Experiment results show that our method improves the performance of solenoid protein recognition substantially.

In microarray data, a bicluster refers to a subset of genes exhibiting consistent patterns over a subset of conditions. In this paper, we propose a method for detecting these biclusters in large gene expression datasets. We consider the bicluster patterns based geometric relations. We use Randomized Hough Transform for sub-bicluster detection in column pair spaces and a spectra graph based combination algorithm is formulated to reduce the time complexity for combining the sub-biclusters. Experiment results demonstrate that our approach reduces the computing time and outperforms existing biclustering algorithms with higher biclustering accuracy.

Protein-ligand interaction is important in drug development and protein design. Previous studies focused on the detection of the potential ligand binding sites but lack of analysis of the true binding sites on the protein. In this work, we use alpha shape models to represent the surface of the protein structure and extract surface patches to study the ligand binding sites. Analysis is performed on X-ray crystallography data of protein-ligand structures. The results show that geometric properties play a significant role in the positioning of ligand molecule in protein-ligand interaction.

Biomarker identification and cancer classification are two closely related problems. In gene expression data sets, the correlation between genes can be high when they share the same biological pathway. Moreover, the gene expression data sets may contain outliers due to either chemical or electrical reasons. A good gene selection method should take group effects into account and be robust to outliers. In this paper, we propose a Laplace naive Bayes model with mean shrinkage (LNB-MS). The Laplace distribution instead of the normal distribution is used as the conditional distribution of the samples for the reasons that it is less sensitive to outliers and has been applied in many fields. The key technique is the L1 penalty imposed on the mean of each class to achieve automatic feature selection. The objective function of the proposed model is a piecewise linear function with respect to the mean of each class, of which the optimal value can be evaluated at the breakpoints simply. An efficient algorithm is designed to estimate the parameters in the model. A new strategy that uses the number of selected features to control the regularization parameter is introduced. Experimental results on simulated data sets and 17 publicly available cancer data sets attest to the accuracy, sparsity, efficiency, and robustness of the proposed algorithm. Many biomarkers identified with our method have been verified in biochemical or biomedical research. The analysis of biological and functional correlation of the genes based on Gene Ontology (GO) terms shows that the proposed method guarantees the selection of highly correlated genes simultaneously

Practical video scene and face recognition systems are sometimes confronted with low-resolution (LR) images. The faces may be very small even if the video is clear, thus it is difficult to directly measure the similarity between the faces and the high-resolution (HR) training samples. Traditional super-resolution (SR) methods based face recognition usually have limited performance because the target of SR may not be consistent with that of classification, and time-consuming SR algorithms are not suitable for real-time applications. In this paper, a new feature extraction method called Coupled Kernel Embedding (CKE) is proposed for LR face recognition without any SR preprocessing. In this method, the final kernel matrix is constructed by concatenating two individual kernel matrices in the diagonal direction, and the (semi-)positively definite properties are preserved for optimization. CKE addresses the problem of comparing multi-modal data that are difficult for conventional methods in practice due to the lack of an efficient similarity measure. Particularly, different kernel types (e.g., linear, Gaussian, polynomial) can be integrated into an uniformed optimization objective, which cannot be achieved by simple linear methods. CKE solves this problem by minimizing the dissimilarities captured by their kernel Gram matrices in the low- and high-resolution spaces. In the implementation, the nonlinear objective function is minimized by a generalized eigenvalue decomposition. Experiments on benchmark and real databases show that our CKE method indeed improves the recognition performance.

In this paper, we present a method to analyze the molecular structure and dynamics of nucleosomes, which are the basic unit in eukaryotic cells. In nucleosomal DNA chains, periodic dinucleotides are key patterns in protein-DNA interactions. This paper aims at revealing the significances of these periodic dinucleotides in the nucleosomal DNA motions. Normal mode analysis (NMA) is used to detect significant structural deformations of nucleosomal DNA. We have found that periodic dinucleotides are usually located at the peaks or valleys of DNA and protein motions, revealing they dominate the nucleosome dynamics. Also, a specific dinucleotide pattern CA/TG appears most frequently.

This paper presents a new approach for short gene recognition in DNA sequences. Three DNA structural features are selected from an analysis of fourteen structural features. The feature values are mapped to new values. Three DNA signals are generated by the three sets of mapped feature values. Then the three DNA signals are normalized and combined into one signal. An auto-regressive (AR) model is used for power spectral density (PSD) estimation of the signal. The experiment result obtained by this method is shown to be comparable to existing exon detection methods which use digital signal processing (DSP). Also the computation complexity of the new method is only 1/3 of that of the method proposed previously.

Microarray gene expression data generally suffers from missing value problem due to a variety of experimental reasons. Since the missing data points can adversely affect downstream analysis, many algorithms have been proposed to impute missing values. In this survey, we provide a comprehensive review of existing missing value imputation algorithms, focusing on their underlying algorithmic techniques and how they utilize local or global information from within the data, or their use of domain knowledge during imputation. In addition, we describe how the imputation results can be validated and the different ways to assess the performance of different imputation algorithms, as well as a discussion on some possible future research directions. It is hoped that this review will give the readers a good understanding of the current development in this field and inspire them to come up with the next generation of imputation algorithms.

The effects of a drug on the genomic scale can be assessed in a three-color cDNA microarray with the three color intensities represented through the so-called hexaMplot. In our recent study, we have shown that the Hough Transform (HT) applied to the hexaMplot can be used to detect groups of coexpressed genes in the normal-disease-drug samples. However, the standard HT is not well suited for the purpose because 1) the assayed genes need first to be hard-partitioned into equally and differentially expressed genes, with HT ignoring possible information in the former group; 2) the hexaMplot coordinates are negatively correlated and there is no direct way of expressing this in the standard HT and 3) it is not clear how to quantify the association of coexpressed genes with the line along which they cluster. We address these deficiencies by formulating a dedicated probabilistic model-based HT. The approach is demonstrated by assessing effects of the drug Rg1 on homocysteine-treated human umbilical vein endothetial cells. Compared with our previous study, we robustly detect stronger natural groupings of coexpressed genes. Moreover, the gene groups show coherent biological functions with high significance, as detected by the Gene Ontology analysis.

Extracting biologically relevant information from DNA microarrays is a very important task for drug development and test, function annotation, and cancer diagnosis. Various clustering methods have been proposed for the analysis of gene expression data, but when analyzing the large and heterogeneous collections of gene expression data, conventional clustering algorithms often cannot produce a satisfactory solution. Biclustering algorithm has been presented as an alternative approach to standard clustering techniques to identify local structures from gene expression data set. These patterns may provide clues about the main biological processes associated with different physiological states. In this paper, different from existing bicluster patterns, we first introduce a more general pattern: correlated bicluster, which has intuitive biological interpretation. Then, we propose a novel transform technique based on singular value decomposition so that identifying correlated-bicluster problem from gene expression matrix is transformed into two global clustering problems. The Mixed-Clustering algorithm and the Lift algorithm are devised to efficiently produce δ-corBiclusters. The biclusters obtained using our method from gene expression data sets of multiple human organs and the yeast Saccharomyces cerevisiae demonstrate clear biological meanings.

This paper presents a new technique for the detection of short exons in DNA sequences. In this method, we analyze the DNA propeller twist and bending stiffness using the autoregressive (AR) model. The linear prediction matrices for the two features are combined to find the same set of linear prediction coefficients, from which we estimate the spectrum of the DNA sequence and detect protein-coding regions based on the 1/3 frequency component. To overcome the non-stationarity of DNA sequences, we use moving windows of different sizes in the AR model. Experiments on the human genome show that our multi-feature based method is superior in performance to existing exon detection algorithms.

Previous studies about protein-DNA interaction focused on the bound structure of DNA-binding proteins and provided good but not practical results. In our work, we apply an alpha shape model to represent the surface structure of the protein-DNA complex and use structural alignment to develop an interface-atom curvature-dependent conditional probability discriminatory function for the prediction of unbound DNA-binding protein. The proposed method provides good performance in predicting unbound structure of DNA-binding protein which is potentially useful in many fields. Computer experiment results show that the curvature-dependent formalism with the optimal parameters can achieve sensitivity ranges from 48.08% to 44.23% and specificity ranges from 73.82% to 84.29%.

This paper investigates the affect of variation of patterns in protein profiles to the identification of disease-specific biomarkers. A correlation-based cluster-space transform is applied to mass spectral data for predicting major adverse cardiac events (MACE). Training and testing data are transformed into cluster spaces by correlation distance based clustering, respectively. Data in the testing cluster that falls into a pair of training clusters is classified by a supervised classifier. Experiment results have shown that proteomic spectra of MACE which vary with certain patterns could be separated by the correlation-based clustering. The cluster-space transform allows better classification accuracy than single-clustered class method for separating disease and healthy samples.

MOTIVATION: Protein-DNA interaction has significant importance in many biological processes. However, the underlying principle of the molecular recognition process is still largely unknown. As more high-resolution 3D structures of protein-DNA complex are becoming available, the surface characteristics of the complex become an important research topic. RESULT: In our work, we apply an alpha shape model to represent the surface structure of the protein-DNA complex and developed an interface-atom curvature-dependent conditional probability discriminatory function for the prediction of protein-DNA interaction. The interface-atom curvature-dependent formalism captures atomic interaction details better than the atomic distance-based method. The proposed method provides good performance in discriminating the native structures from the docking decoy sets, and outperforms the distance-dependent formalism in terms of the z-score. Computer experiment results show that the curvature-dependent formalism with the optimal parameters can achieve a native z-score of -8.17 in discriminating the native structure from the highest surface-complementarity scored decoy set and a native z-score of -7.38 in discriminating the native structure from the lowest RMSD decoy set. The interface-atom curvature-dependent formalism can also be used to predict apo version of DNA-binding proteins. These results suggest that the interface-atom curvature-dependent formalism has a good prediction capability for protein-DNA interactions. AVAILABILITY: The code and data sets are available for download on http://www.hy8.com/bioinformatics.htm CONTACT: [email protected].

BACKGROUND: Gene clustering for annotating gene functions is one of the fundamental issues in bioinformatics. The best clustering solution is often regularized by multiple constraints such as gene expressions, Gene Ontology (GO) annotations and gene network structures. How to integrate multiple pieces of constraints for an optimal clustering solution still remains an unsolved problem. RESULTS: We propose a novel multiconstrained gene clustering (MGC) method within the generalized projection onto convex sets (POCS) framework used widely in image reconstruction. Each constraint is formulated as a corresponding set. The generalized projector iteratively projects the clustering solution onto these sets in order to find a consistent solution included in the intersection set that satisfies all constraints. Compared with previous MGC methods, POCS can integrate multiple constraints from different nature without distorting the original constraints. To evaluate the clustering solution, we also propose a new performance measure referred to as Gene Log Likelihood (GLL) that considers genes having more than one function and hence in more than one cluster. Comparative experimental results show that our POCS-based gene clustering method outperforms current state-of-the-art MGC methods. CONCLUSIONS: The POCS-based MGC method can successfully combine multiple constraints from different nature for gene clustering. Also, the proposed GLL is an effective performance measure for the soft clustering solutions.

This paper integrates the signal, context, and structure features for genome-wide human promoter recognition, which is important in improving genome annotation and analyzing transcriptional regulation without experimental supports of ESTs, cDNAs, or mRNAs. First, CpG islands are salient biological signals associated with approximately 50 percent of mammalian promoters. Second, the genomic context of promoters may have biological significance, which is based on n-mers (sequences of n bases long) and their statistics estimated from training samples. Third, sequence-dependent DNA flexibility originates from DNA 3D structures and plays an important role in guiding transcription factors to the target site in promoters. Employing decision trees, we combine above signal, context, and structure features to build a hierarchical promoter recognition system called SCS. Experimental results on controlled data sets and the entire human genome demonstrate that SCS is significantly superior in terms of sensitivity and specificity as compared to other state-of-the-art methods. The SCS promoter recognition system is available online as supplemental materials for academic use and can be found on the Computer Society Digital Library at http://doi.ieeecomputersociety.org/10.1109/TCBB.2008.95.

Support vector regression and regularization networks are kernel-based techniques for solving the regression problem of recovering the unknown function from sample data. The choice of the kernel function, which determines the mapping between the input space and the feature space, is of crucial importance to such learning machines. Estimating the irregular function with a multiscale structure that comprises both the steep variations and the smooth variations is a hard problem. The result achieved by the traditional Gaussian kernel is often unsatisfactory, because it cannot simultaneously avoid underfitting and overfitting. In this paper, we present a new class of kernel functions derived from the framelet system. A framelet is a tight wavelet frame constructed via multiresolution analysis and has the merit of both wavelets and frames. The construction and approximation properties of framelets have been well studied. Our goal is to combine the power of framelet representation with the merit of kernel methods on learning from sparse data. The proposed framelet kernel has the ability to approximate functions with a multiscale structure and can reduce the influence of noise in data. Experiments on both simulated and real data illustrate the usefulness of the new kernels.

Many cellular processes exhibit cyclic behaviors. Hence, one important task in gene expression data analysis is to detect subset of genes that exhibit periodicity in their gene expression time series profiles. Unfortunately, gene expression time series profiles are usually of very short length, with very few periods, unevenly sampled, and are highly contaminated with noise. This makes detection of periodic profiles a very challenging problem. In this paper, we present several effective computational techniques developed recently in our research group for the reliable detection of short periodic gene expression time series profiles.

This review describes important advances that have been made during the past decade for genome-wide human promoter recognition. Interest in promoter recognition algorithms on a genome-wide scale is worldwide and touches on a number of practical systems that are important in analysis of gene regulation and in genome annotation without experimental support of ESTs, cDNAs or mRNAs. The main focus of this review is on feature extraction and model selection for accurate human promoter recognition, with descriptions of what they are, what has been accomplished, and what remains to be done.

This paper examines the classification capability of different Gabor representations for human face recognition. Usually, Gabor filter responses for eight orientations and five scales for each orientation are calculated and all 40 basic feature vectors are concatenated to assemble the Gabor feature vector. This work explores 70 different Gabor feature vector extraction techniques for face recognition. The main goal is to determine the characteristics of the 40 basic Gabor feature vectors and to devise a faster Gabor feature extraction method. Among all the 40 basic Gabor feature representations the filter responses acquired from the largest scale at smallest relative orientation change (with respect to face) shows the highest discriminating ability for face recognition while classification is performed using three classification methods: probabilistic neural networks (PNN), support vector machines (SVM) and decision trees (DT). A 40 times faster summation based Gabor representation shows about 98% recognition rate while classification is performed using SVM. In this representation all 40 basic Gabor feature vectors are summed to form the summation based Gabor feature vector. In the experiment, a sixth order data tensor containing the basic Gabor feature vectors is constructed, for all the operations.

Missing value estimation is important in DNA microarray data analysis. A number of algorithms have been developed to solve this problem, but they have several limitations. Most existing algorithms are not able to deal with the situation where a particular time point (column) of the data is missing entirely. In this paper, we present an autoregressive-model-based missing value estimation method (ARLSimpute) that takes into account the dynamic property of microarray temporal data and the local similarity structures in the data. ARLSimpute is especially effective for the situation where a particular time point contains many missing values or where the entire time point is missing. Experiment results suggest that our proposed algorithm is an accurate missing value estimator in comparison with other imputation methods on simulated as well as real microarray time series datasets.

Tandem repeats, which occur frequently in genomes, are related to gene regulatory functions and various diseases. In this paper, an efficient algorithm for tandem repeat detection is proposed. In our method, the spectrogram of a DNA sequence is analyzed based on the autoregressive model. Then, significant peaks in the spectrogram are selected, and the corresponding regions in the DNA sequence are analyzed to search for tandem repeats. Experiment results show that our method has a superior performance in comparison with other algorithms.

The analysis of gene expression data of breast cancer is important for discovering the signatures that can classify different subtypes of tumors and predict prognosis. Biclustering algorithms have been proven to be able to group the genes with similar expression patterns under a number of samples and offer the capability to analyze the microarray data of cancer. In this study, we propose a new biclustering algorithm which uses an evolutionary search procedure. The algorithm is applied to the conditions to search for combinations of conditions for a potential bicluster. Preliminary results using synthetic and real yeast data sets demonstrate that our algorithm outperforms several existing ones. We have also applied the method to real microarray data sets of breast cancer, and successfully found several biclusters, which can be used as signatures for differentiating tumor types.

Fuzzy biclustering analysis is a useful tool for identifying relevant subsets of microarray data. This paper proposes a fuzzy biclustering clustering method for microarray data analysis. The method employs a combination of the Nelder-Mead and min-max algorithm to construct hierarchically structured biclustering. The method can automatically identify the groups of genes that show similar expression patterns under a specific subset of the samples.

High dimensionality, noisy features and outliers can cause problems in cluster analysis. Many existing methods can handle one of the problems well but not the others. In this paper, we propose a new clustering algorithm to solve these problems. The basic idea is to control the support of the optimization procedure so that the effect produced by those contaminated samples and dimensions is greatly reduced. This is achieved by using multiple supports. Initially, a large support is used and then its size is reduced and eventually only a subgroup of data samples is considered for clustering. This procedure can filter out lots of contaminated information. Experiment results show that the proposed method effectively resolves all these problems. It outperforms existing ones for real world high dimensional datasets.

In this paper, we propose to combine linear prediction coefficients from the autoregressive model (AR) and the time series itself as features for the clustering algorithm. The purpose of the use of the AR model is to realize the importance of dynamic modeling of microarray time series data. We define the distance among the time series profiles using the autoregressive model and use the hierarchical clustering and the k-means clustering methods for comparison. The results show that the performance of the clustering DNA microarray time course profile is increased with the linear prediction coefficients in addition to the time series itself used as features.

A subset of genes sharing compatible expression patterns under a subset of conditions can be found from DNA microarray data using biclustering algorithms. In this paper, we present a novel geometrical biclustering algorithm in combination with gene ontology annotations to identify the gene functional biclusters. Unlike many existing biclustering algorithms, we first consider the biclustering patterns through geometrical interpretation. Such a perspective makes it possible to unify the formulation of different types of biclusters as hyperplanes in spatial space and facilitates the use of a generic plane finding algorithm for bicluster detection. In our bottom-up biclustering algorithm, the well-known Hough transform is first employed in pair-column spaces to reduce the computation complexity and then the resulting patterns are merged step by step into large-size biclusters incorporated with gene functional modules. The algorithm integrates the numerical characteristics in a gene expression matrix and the gene functions in the biological activities. Our experiments on real data show that the new algorithm outperforms most existing methods for mining gene functional biclusters.

BACKGROUND: In DNA microarray experiments, discovering groups of genes that share similar transcriptional characteristics is instrumental in functional annotation, tissue classification and motif identification. However, in many situations a subset of genes only exhibits consistent pattern over a subset of conditions. Conventional clustering algorithms that deal with the entire row or column in an expression matrix would therefore fail to detect these useful patterns in the data. Recently, biclustering has been proposed to detect a subset of genes exhibiting consistent pattern over a subset of conditions. However, most existing biclustering algorithms are based on searching for sub-matrices within a data matrix by optimizing certain heuristically defined merit functions. Moreover, most of these algorithms can only detect a restricted set of bicluster patterns. RESULTS: In this paper, we present a novel geometric perspective for the biclustering problem. The biclustering process is interpreted as the detection of linear geometries in a high dimensional data space. Such a new perspective views biclusters with different patterns as hyperplanes in a high dimensional space, and allows us to handle different types of linear patterns simultaneously by matching a specific set of linear geometries. This geometric viewpoint also inspires us to propose a generic bicluster pattern, i.e. the linear coherent model that unifies the seemingly incompatible additive and multiplicative bicluster models. As a particular realization of our framework, we have implemented a Hough transform-based hyperplane detection algorithm. The experimental results on human lymphoma gene expression dataset show that our algorithm can find biologically significant subsets of genes. CONCLUSION: We have proposed a novel geometric interpretation of the biclustering problem. We have shown that many common types of bicluster are just different spatial arrangements of hyperplanes in a high dimensional data space. An implementation of the geometric framework using the Fast Hough transform for hyperplane detection can be used to discover biologically significant subsets of genes under subsets of conditions for microarray data analysis.

BACKGROUND: Identification of differentially expressed genes is a typical objective when analyzing gene expression data. Recently, Bayesian hierarchical models have become increasingly popular to solve this type of problems. These models show good performance in accommodating noise, variability and low replication of microarray data. However, the correlation between different fluorescent signals measured from a gene spot is ignored, which can diversely affect the data analysis step. In fact, the intensities of the two signals are significantly correlated across samples. The larger the log-transformed intensities are, the smaller the correlation is. RESULTS: Motivated by the complicated error relations in microarray data, we propose a multivariate hierarchical Bayesian framework for data analysis in the replicated microarray experiments. Gene expression data are modelled by a multivariate normal distribution, parameterized by the corresponding mean vectors and covariance matrixes with a conjugate prior distribution. Within the Bayesian framework, a generalized likelihood ratio test (GLRT) is also developed to infer the gene expression patterns. Simulation studies show that the proposed approach presents better operating characteristics and lower false discovery rate (FDR) than existing methods, especially when the correlation coefficient is large. The approach is illustrated with two examples of microarray analysis. The proposed method successfully detects significant genes closely related to the experimental states, which are verified by the biological information. CONCLUSIONS: The multivariate Bayesian model, compatible with the dependence between mean and variance in the univariate Bayesian model, relaxes the constant coefficient of variation assumption between measurements by adding a covariance structure. This model improves the identification of differentially expressed genes significantly since the Bayesian model fit well with the microarray data.

High-dimensional data and the small sample size problem occur in many modern pattern classification applications such as face recognition and gene expression data analysis. To deal with such data, one important step is dimensionality reduction. Principal component analysis (PCA) and between-group analysis (BGA) are two commonly used methods, and various extensions of these two methods exist. The principle of these two approaches comes from their best approximation property. From a pattern recognition perspective, we show that PCA, which is based on the total scatter matrix, preserves linear separability, and BGA, which is based on between-class scatter matrix, retains the distance between class centroids. Moreover, we propose an automatic nonparameter uncorrelated discriminant analysis (UDA) algorithm based on the maximum margin criterion (MMC). The extracted features via UDA are statistically uncorrelated. UDA combines rank-preserving dimensionality reduction and constraint discriminant analysis and also serves as an effective solution for the small-sample-size problem. Experiments with face images and gene expression data sets are conducted to evaluate UDA in terms of classification accuracy and robustness.

In this paper, a method is proposed for detecting blood vessels in pathological retina images. In the proposed method, blood vessel-like objects are extracted using the Laplacian operator and noisy objects are pruned according to the centerlines, which are detected using the normalized gradient vector field. The method has been tested with all the pathological retina images in the publicly available STARE database. Experiment results show that the method can avoid detecting false vessels in pathological regions and can produce reliable results for healthy regions.

This paper presents an approach for image recovery from broken image streams based on an image continuity model. The information to be recovered includes the header of an image (width, height, color space etc.) and the data stream which might have been partially damaged. We begin our discussion on common image formats and necessary information for the image recovery. We then define several measures on image contents to facilitate the image recovery. Using these measures, an effective recovery algorithm is proposed. Experimental results show that the algorithm can successfully recover images of different formats, even multi-frame images from broken image streams in most cases.

In this paper, we propose a new algorithm for classification of multi-dimensional data, in which noisy features are distributed in different dimensions of different groups. This kind of datasets violate the assumption of many existing dimension reduction methods, which assume all the groups have the noisy features in the same dimensions and the pruning operation is conducted on the same dimensions of all the groups. Our strategy to resolve this problem is to use multi-classifiers. Each classifier engages different set of dimensions and carries out dimensionality reduction separately. Experiment results on six real world data sets show that the proposed algorithm has a superior to existing ones.

UNLABELLED: Repetitive DNA sequences play paramount biological roles, such as gene variation and regulatory functions on gene expressions. Until now, detection of various kinds of DNA repeats accurately is still an open problem. In this article, we propose a new method and a visualization tool for detecting DNA repeats in a 2D plane of location and frequency by using optimized moving window spectral analysis. The spectrogram can display the general distribution of repetitive sequences while showing the repeat period, length and location without any prior knowledge. Experimental results demonstrate that our method is accurate and robust even under the condition of excessive mutating and interleaving. AVAILABILITY: Available on http://www.hy8.com/~tec/sw01/omwsa01.zip. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Periodogram analysis of time-series is widespread in biology. A new challenge for analyzing the microarray time series data is to identify genes that are periodically expressed. Such challenge occurs due to the fact that the observed time series usually exhibit non-idealities, such as noise, short length, and unevenly sampled time points. Most methods used in the literature operate on evenly sampled time series and are not suitable for unevenly sampled time series. RESULTS: For evenly sampled data, methods based on the classical Fourier periodogram are often used to detect periodically expressed gene. Recently, the Lomb-Scargle algorithm has been applied to unevenly sampled gene expression data for spectral estimation. However, since the Lomb-Scargle method assumes that there is a single stationary sinusoid wave with infinite support, it introduces spurious periodic components in the periodogram for data with a finite length. In this paper, we propose a new spectral estimation algorithm for unevenly sampled gene expression data. The new method is based on signal reconstruction in a shift-invariant signal space, where a direct spectral estimation procedure is developed using the B-spline basis. Experiments on simulated noisy gene expression profiles show that our algorithm is superior to the Lomb-Scargle algorithm and the classical Fourier periodogram based method in detecting periodically expressed genes. We have applied our algorithm to the Plasmodium falciparum and Yeast gene expression data and the results show that the algorithm is able to detect biologically meaningful periodically expressed genes. CONCLUSION: We have proposed an effective method for identifying periodic genes in unevenly sampled space of microarray time series gene expression data. The method can also be used as an effective tool for gene expression time series interpolation or resampling.

BACKGROUND: Three-color microarray experiments can be performed to assess drug effects on the genomic scale. The methodology may be useful in shortening the cycle, reducing the cost, and improving the efficiency in drug discovery and development compared with the commonly used dual-color technology. A visualization tool, the hexaMplot, is able to show the interrelations of gene expressions in normal-disease-drug samples in three-color microarray data. However, it is not enough to assess the complicated drug therapeutic effects based on the plot alone. It is important to explore more effective tools so that a deeper insight into gene expression patterns can be gained with three-color microarrays. RESULTS: Based on the celebrated Hough transform, a novel algorithm, HoughFeature, is proposed to extract line features in the hexaMplot corresponding to different drug effects. Drug therapy results can then be divided into a number of levels in relation to different groups of genes. We apply the framework to experimental microarray data to assess the complex effects of Rg1 (an extract of Chinese medicine) on Hcy-related HUVECs in details. Differentially expressed genes are classified into 15 functional groups corresponding to different levels of drug effects. CONCLUSION: Our study shows that the HoughFeature algorithm can reveal natural cluster patterns in gene expression data of normal-disease-drug samples. It provides both qualitative and quantitative information about up- or down-regulated genes. The methodology can be employed to predict disease susceptibility in gene therapy and assess drug effects on the disease based on three-color microarray data.

In this paper, we propose spectral analysis method to identify periodically expressed genes in microarray data using a forward and backward linear prediction (FBLP) model and the singular value decomposition (SVD) (FBLP-SVD) algorithm. The spectrum-mean-subtraction method is employed prior to this analysis as a pre-filtering procedure. The combination of the spectrum-mean-subtraction and FBLP-SVD algorithm offers a effective tool for periodicity identification. Using our technique, more genes have been successful identified as periodic genes in the genome of Saccharomyces cerevisiae.

Given a newly found gene of some particular genome and a database of sequences whose functions have been known, it must be very helpful if we can search through the database and identify those that are similar to the particular new sequence. The search results may help us to understand the functional role, regulation, and expression of the new gene by the inference from the similar database sequences. This is the task of any methods developed for biological database searching. In this paper we present a new application of the theories of linear predictive coding, vector quantization, and hidden Markov models to address the problem of DNA sequence similarity search where there is no need for sequence alignment. The proposed approach has been tested and compared with some existing methods against real DNA and genomic datasets. The experimental results demonstrate its potential use for such purpose.

A tandem repeat (TR) is a DNA sequence where a pattern of nucleotides is repeated a number of times. TRs cover more than ten percent of the human genome. They have been proven to play an important role in human diseases, regulation, and evolution. TRs vary for different individuals, so they are commonly used in human gene mapping, linkage studies, and forensic DNA fingerprinting analysis. In this paper, an efficient algorithm is presented for detecting TRs, especially short tandem repeats (STRs), in a DNA sequence. The algorithm, based on the autoregressive (AR) model, is to analyze the spectrum of the DNA sequences. Our algorithm can find TRs effectively and quickly. Furthermore, it is robust to mutations, deletions, and insertions. In comparison with the fast Fourier transform (FFT), our results show that the AR model based algorithm can provide more detailed qualitative information than the FFT when we analyze the spectrum of the STRs. Here, the methods and ideas underlying the algorithm are presented and the effectiveness of the algorithm on TRs is demonstrated.

MOTIVATION: Promoter prediction is important for the analysis of gene regulations. Although a number of promoter prediction algorithms have been reported in literature, significant improvement in prediction accuracy remains a challenge. In this paper, an effective promoter identification algorithm, which is called PromoterExplorer, is proposed. In our approach, we analyze the different roles of various features, that is, local distribution of pentamers, positional CpG island features and digitized DNA sequence, and then combine them to build a high-dimensional input vector. A cascade AdaBoost-based learning procedure is adopted to select the most 'informative' or 'discriminating' features to build a sequence of weak classifiers, which are combined to form a strong classifier so as to achieve a better performance. The cascade structure used for identification can also reduce the false positive. RESULTS: PromoterExplorer is tested based on large-scale DNA sequences from different databases, including the EPD, DBTSS, GenBank and human chromosome 22. Experimental results show that consistent and promising performance can be achieved.

A different contour search algorithm is presented in this paper that provides a faster convergence to the object contours than both the greedy snake algorithm (GSA) and the fast greedy snake (FGSA) algorithm. This new algorithm performs the search in an alternate skipping way between the even and odd nodes (snaxels) of a snake with different step sizes such that the snake moves to a likely local minimum in a twisting way. The alternative step sizes are adjusted so that the snake is less likely to be trapped at a pseudo-local minimum. The iteration process is based on a coarse-to-fine approach to improve the convergence. The proposed algorithm is compared with the FGSA algorithm that employs two alternating search patterns without altering the search step size. The algorithm is also applied in conjunction with the subband decomposition to extract face profiles in a hierarchical way.

Several cluster validity measures have been proposed for evaluating clustering results. However, existing methods may not work well for the following two kinds of data sets. The first one is that the data set contains cluster groups with different densities. The second one is that some of the cluster groups are closely positioned. In this paper, we introduce a new cluster validity index. In this method, we define the index as the ratio between the squared total length of the data eigen-axes and the between-cluster separation. Compared with existing cluster validity indices, the proposed index produces more accurate results and is able to handle the two kinds of data sets mentioned above.

It is well known that low bit rate block-based discrete cosine transform coded image exhibits visually annoying coding artifacts. In this paper, we proposed a projection onto convex sets (PCOS)-based deblocking algorithm using a novel region smoothness constraint set for graphic images containing objects with smooth regions. The smoothness constraint set is obtained by an explicit modeling of smooth regions in the image using a spatially adaptive thin-plate spline. In contrast to most deblocking algorithms which enforce smoothness just around the 8/spl times/8 block boundaries, our algorithm enforces smoothness in regions which could possibly span several blocks. We showed that convergence of our algorithm could be reached within one iteration. The performance of the proposed algorithm is evaluated visually and quantitatively in term of peak signal-to-noise ratios and the mean squared difference of slope metric, which measures the impact of the blocking effects, for several graphic images. The results show that our algorithm can effectively suppress blockiness in smooth regions while still preserving the sharpness of object edges.

A deblocking method based on projection onto convex sets (POCS) is proposed to reduce blocking artifacts in compressed images coded by the block discrete cosine transform. The method differs from existing POCS-based methods in three aspects. Firstly, the adjustment of a pixel's intensity is determined by local properties of the pixel. Secondly, three locally adaptive constraint sets are introduced to improve deblocking results. Thirdly, the human visual system modeling and relaxed projections are incorporated to make pixels adjust appropriately. The method is tested on typical images with excellent results.

An image segmentation algorithm based on adaptive fuzzy c-means (FCM) clustering is presented in this paper. In the conventional FCM clustering algorithm, cluster assignment is based solely on the distribution of pixel attributes in the feature space, and does not take into consideration the spatial distribution of pixels in an image. By introducing a novel dissimilarity index in the modified FCM objective function, the new adaptive fuzzy clustering algorithm is capable of utilizing local contextual information to impose local spatial continuity, thus exploiting the high inter-pixel correlation inherent in most real-world images. The incorporation of local spatial continuity allows the suppression of noise and helps to resolve classification ambiguity. To account for smooth intensity variation within each homogenous region in an image, a multiplicative field is introduced to each of the fixed FCM cluster prototype. The multiplicative field effectively makes the fixed cluster prototype adaptive to slow smooth within-cluster intensity variation, and allows homogenous regions with slow smooth intensity variation to be segmented as a whole. Experimental results with synthetic and real color images have shown the effectiveness of the proposed algorithm.

We present a method for performing face recognition based on the fractal neighbor distance (FND). The FND has previously been used for face recognition. What distinguishes our method from others is that we incorporate the use of localized weights with the FND. In a local-to-global feature matching approach, a set of localized weights is used with an algorithm based on the FND that searches for local features. A global score is then derived from each localized score. This set of weights is designed to concentrate around the eyes and nose region of the face, because they contain more discriminating features.

In a constrained total least squares algorithm (CTLS), the selection of a minimal algebraic set of linearly independent random variables to express the noise matrix /spl Delta/C is an important task. A fast algorithm is provided using the possibly dependent random variables set. We showed that it can be viewed as a combination of the CTLS method of V.Z. Mesarovic et al. (see IEEE Trans. Image Process., vol.4, p.1096-107, 1995) and the RLS method when the noise is Gaussian. Our experimental study indicates that our algorithm has better visual and objective quality, while having a much lower computation cost. Moreover, our algorithm can also handle a more general noise model.

Human facial animation is an interesting and difficult problem in computer graphics. In this paper, a novel B-spline (NURBS) muscle system is proposed to simulate a 3D facial expression and talking animation. The system gets the lip shape parameters from the video, which captures a real person's lip movement, to control the proper muscles to form different phonemes. The muscles are constructed by the non-uniform rational B-spline curves, which are based on anatomical knowledge. By using different number of control points on the muscles, more detailed facial expression and mouth shapes can be simulated. We demonstrate the flexibility of our model by simulating different emotions and lip-sync to a video with a talking head using the automatically extracted lip parameters.

MOTIVATION: Microarray time-series data provides us a possible means for identification of transcriptional regulation relationships among genes. Currently, the most commonly used method in determining whether or not two genes have a potential regulatory relationship is to measure their expressional similarity using Pearson's correlation coefficient. Although this traditional correlation method has been successfully applied to find functionally correlated genes, it does have many limitations. In the hope of overcoming such circumstances and getting more insights into the transcriptional regulatory issue, we propose an autoregressive (AR)-based technique for detection of potential regulated gene pairs from time-series microarray measurements. RESULTS: We use the well-known AR modeling technique to characterize temporal gene expression data from the Spellman's alpha-synchronized yeast cell-cycle experiment. In this method, time-series expression profiles are decomposed into spectral components and correlations between profiles are then computed in a component-wise sense. We show how these component-wise correlations reveal possible regulatory relationships. Our technique is applied on known transcriptional regulations and is able to identify many of those missed by the traditional correlation method.

It is well known that at low-bit-rate block discrete cosine transform compressed image exhibits visually annoying blocking and ringing artifacts. In this paper, we propose a noniterative, wavelet-based deblocking algorithm to reduce both types of artifacts. The algorithm exploits the fact that block discontinuities are constrained by the dc quantization interval of the quantization table, as well as the behavior of wavelet modulus maxima evolution across wavelet scales to derive appropriate threshold maps at different wavelet scales. Since ringing artifacts occur near strong edges, which can be located either along block boundaries or within blocks, suppression of block discontinuities does not always reduce ringing artifacts. By exploiting the behavior of ringing artifacts in the wavelet domain, we propose a simple yet effective method for the suppression of such artifacts. The proposed algorithm can suppress both block discontinuities and ringing artifacts effectively while preserving true edges and textural information. Simulation results and extensive comparative study with both iterative and noniterative methods reported in the literature have shown the effectiveness of our algorithm.

Cluster analysis of gene expression data from a cDNA microarray is useful for identifying biologically relevant groups of genes. However, finding the natural clusters in the data and estimating the correct number of clusters are still two largely unsolved problems. In this paper, we propose a new clustering framework that is able to address both these problems. By using the one-prototype-take-one-cluster (OPTOC) competitive learning paradigm, the proposed algorithm can find natural clusters in the input data, and the clustering solution is not sensitive to initialization. In order to estimate the number of distinct clusters in the data, we propose a cluster splitting and merging strategy. We have applied the new algorithm to simulated gene expression data for which the correct distribution of genes over clusters is known a priori. The results show that the proposed algorithm can find natural clusters and give the correct number of clusters. The algorithm has also been tested on real gene expression changes during yeast cell cycle, for which the fundamental patterns of gene expression and assignment of genes to clusters are well understood from numerous previous studies. Comparative studies with several clustering algorithms illustrate the effectiveness of our method.

The fuzzy curve-tracing (FCT) algorithm can be used to extract a smooth curve from unordered noisy data. In this paper, we analyze the convergence property of the algorithm based on the diagonal dominance requirement of the matrix used in the clustering procedure and prove that the algorithm is guaranteed to converge if the weighting coefficient for the smoothness constraint is chosen properly. Based on the convergence condition, we develop several methods for fast and reliable implementation of the algorithm. We show that the algorithm can be initialized with a user-defined curve in many cases, that a multiresolution clustering based approach and an image down-sampling scheme can be used to improve the algorithm stability and speed and that two types of traps can be removed to correct the mistakes in curve tracing. We demonstrate several advantages of our algorithm over the commonly used snake models for boundary detection and several methods for principle curve extraction.

A fuzzy c-means based adaptive clustering algorithm is proposed for the fuzzy segmentation of 3D MR brain images, which are typically corrupted by noise and intensity non-uniformity (INU) artifact. The proposed algorithm enforces the spatial continuity constraint to account for the spatial correlations between image voxels, resulting in the suppression of noise and classification ambiguity. The INU artifact is compensated for by the introduction of a pseudo-3D bias field, which is modeled as a stack of smooth B-spline surfaces with continuity enforced across slices. The efficacy of the proposed algorithm is demonstrated experimentally using both simulated and real MR images.

Blocking artifacts often appear in compressed images coded by the block discrete cosine transform. The paper presents a smoothing method for reducing those artifacts. The method is based on adjusting pixel intensities of a blocky image to values determined by a smooth model image. Certain points in the blocky image are selected to construct a triangular mesh. The mesh forms the model image. The adjusted image is further subject to a narrow quantization constraint. The method produces superior images in terms of subjective quality and objective quality when compared to existing methods. It is fast, since no iterations are required.

An adaptive spatial fuzzy c-means clustering algorithm is presented in this paper for the segmentation of three-dimensional (3-D) magnetic resonance (MR) images. The input images may be corrupted by noise and intensity nonuniformity (INU) artifact. The proposed algorithm takes into account the spatial continuity constraints by using a dissimilarity index that allows spatial interactions between image voxels. The local spatial continuity constraint reduces the noise effect and the classification ambiguity. The INU artifact is formulated as a multiplicative bias field affecting the true MR imaging signal. By modeling the log bias field as a stack of smoothing B-spline surfaces, with continuity enforced across slices, the computation of the 3-D bias field reduces to that of finding the B-spline coefficients, which can be obtained using a computationally efficient two-stage algorithm. The efficacy of the proposed algorithm is demonstrated by extensive segmentation experiments using both simulated and real MR images and by comparison with other published algorithms.

A novel protection onto convex sets (POCS) method is presented for the suppression of blocking and ringing artifacts in a compressed image that contains homogeneous regions. A new family of convex smoothness constraint sets is introduced, using the uniformity property of image regions. This set of constraints allows different degrees of smoothing in different regions of the image, while preserving the image edges. The regions are segmented using the fuzzy c-means algorithm, which allows ambiguous pixels to be left unclassified. Experimental results on JPEG compressed images demonstrate that the proposed algorithm yields visually superior images compared to several of the previously reported POCS deblocking algorithms for the class of images considered.

Computer animation has come a long way during the last decade and is now capable of producing near-realistic rendered 3D computer graphics models of expressive, talking, acting humanoids and other characters inhabiting virtual worlds. However, the component of work that needs to be done by animators and artists in producing these synthetic character performances is quite significant. In this paper, we present an expert system based on fuzzy knowledge bases that helps in moving towards automating the task of animating virtual human heads and faces. Our Virtual Actor (Vactor) framework is based on several subsystems that use mainly fuzzy and some non-fuzzy rules to teach virtual actors to know the emotions and gestures to use in different situations. Theories of emotion, personality, dialogue, and acting, as well as empirical evidence are incorporated into our framework and knowledge bases to produce convincing results.

This paper presents an approach to region feature extraction and region matching for cartoon image processing. By introducing the inertia coordinate system, the invariant features of regions have been extracted. Fuzzy dissimilarity is proposed as a matching variable. To simplify the optimal region matching, an optimal matching pair theorem is proposed and proved. Experiments are conducted for verifying the feature extraction and the optimal matching pair theorem.

This paper presents a non-pixel-based skeletonization method for vectorizing cartoon images. The constrained Delaunay triangulation technique is applied to subdivide a shape into a set of non-overlapping triangles. Then, certain triangles in the triangulation are merged to remove artifacts. The skeleton of a shape is obtained from the skeletons of its constituent parts. Experimental results show that the proposed method is more accurate and efficient than a typical thinning method.

An adaptive, spatially constrained fuzzy clustering algorithm for image segmentation is presented. By using a novel dissimilarity index in the cost function, our fuzzy clustering algorithm is capable of utilising local contextual information in a 3/spl times/3 neighborhood to impose local spatial continuity, thus exploiting the high inter-pixel correlation inherent in most realworld images. This has the effects of smoothing out random noise and resolving classification ambiguities. By introducing a multiplicative bias field to the fixed cluster prototypes, the cluster prototypes effectively become adaptive to nonstationarity in the image intensity. This allows non-planar regions or objects to be segmented meaningfully. Experimental results have shown the effectiveness of the proposed fuzzy clustering algorithm.

This paper presents an approach to object feature extraction and object matching for line object processing. The invariant features of objects have been extracted based on inertia coordinate systems. Weighted fuzzy dissimilarity is proposed as a matching variable. To simplify the optimal object matching, an optimal matching pair theorem is proposed. Experiments are conducted for verifying the feature extraction and optimal matching pair theorem.

This paper presents an algorithm designed to segment and classify newspaper documents. A notable feature of this algorithm is the ability to detect lines in the document - including lines that are connected to other components. A bottom-up approach is used to segment the image into patterns, and then each pattern is classified into one of seven types. Complete regions are then formed from the classified patterns.

Artistic documents often contain text along curved paths. Commonly used computer algorithms for text extraction, which only deal with text along straight lines, cannot be employed to analyze these artistic documents. In this paper, we solve this problem using the fuzzy curve-tracing algorithm. In our method, the character pixels are grouped based on the fuzzy c-means algorithm to reduce the amount of data. Then the cluster centers are connected to form the initial curve representing the text path. Finally the character pixels are clustered again under the constraint that the path must be smooth. Results from several experiments are presented to show the effectiveness of our method.

This paper introduces a robust topology-adaptive snake to extend the topological adaptability and flexibility of "snakes". Based on the attractable snake model, three embedded schemes, the robust self-looping process scheme, efficient contour-merging scheme and adaptive interpolation scheme, are proposed. The new snake model is able to: evolve consistently towards its target objects, handle topological changes (i.e., splitting or merging) automatically when necessary and conform to more complicated geometries and topologies, without restrictive requirements on the initial conditions of the snake model (including its parameter settings and its initial contour status) or on its deformation movement. The experiment results using the proposed model for various images are presented.

Digital watermarking is widely believed to be a valid means to discourage illicit distribution of information content. Digital watermarking methods for text documents are limited because of the binary nature of text documents. A distinct feature of a text document is its space patterning. We propose a new approach in text watermarking in which interword spaces of different text lines are slightly modified. After the modification, the average spaces of various lines have the characteristics of a sine wave and the wave constitutes a mark. Both nonblind and blind watermarking algorithms are discussed. Preliminary experiments have shown promising results. Our experiments suggest that space patterning of text documents can be a useful tool in digital watermarking.

A new post-processing method is proposed in the wavelet domain for the suppression of blocking artifacts in compressed images. The novelty of our method is that we can obtain soft-threshold values based on the difference between the wavelet transform coefficients of the image blocks and the coefficients of the entire image to threshold high-frequency wavelet coefficients in different subbands using different values and strategies. The threshold value is made adaptive to different images and characteristics of blocking artifacts. In particular the new method is robust, fast, and works remarkably well for different discrete cosine transform-based compressed images at low bit rates. The method is nonlinear, computationally efficient, and spatially adaptive. Another advantage of the new method is that it retains sharp features in the images since it only removes artifacts. Experimental results show that the proposed method can achieve a significantly improved visual quality and also increase the PSNR in the output image. Our method also performs better than algorithms based on overcomplete wavelet presentation for images containing a large portion of texture.

Thresholding is a commonly used technique in image segmentation. Selecting the correct thresholds is a critical issue. In this paper, the relationship between a probability partition (PP) and a fuzzy c-partition (FP) in thresholding is given. This relationship and the entropy approach are used to derive a thresholding technique to select the best fuzzy c-partition. The measure of the selection quality is the compatibility between the FP and the PP generated by the problem. An entropy function defined by the PP and FP is used to measure the compatibility. A necessary condition of the entropy function arriving at a maximum is derived. Based on this condition, an efficient algorithm for three-level thresholding is deduced. Experiments to verify the efficiency of the proposed method and comparison to some existing techniques are also presented. The experiment results show that our proposed method gives the best performance in three-level thresholding using fuzzy c-partition.

This paper presents a fuzzy clustering algorithm for the extraction of a smooth curve from unordered noisy data. In this method, the input data are first clustered into different regions using the fuzzy c-means algorithm and each region is represented by its cluster center. Neighboring cluster centers are linked to produce a graph according to the average class membership values. Loops in the graph are removed to form a curve according to spatial relations of the cluster centers. The input samples are then reclustered using the fuzzy c-means (FCM) algorithm, with the constraint that the curve must be smooth. The method has been tested with both open and closed curves with good results.

A major problem with traditional skeletonization algorithms is that their results do not always conform to human perceptions since they often contain unwanted artifacts. This paper presents an indirect skeletonization method to reduce these artifacts. The method is based on analyzing regularities and singularities of shapes. A shape is first partitioned into a set of triangles using the constrained Delaunay triangulation technique. Then, regular and singular regions of the shape are identified from the partitioning. Finally, singular regions are stabilized to produce a better result. Experiments show that skeletons obtained from the proposed method closely resemble human perceptions of the underlying shapes.

Fractal image coding, which has been used successfully for image compression, has previously found applications in other areas of image processing, such as object recognition, segmentation and contour extraction. The parameters of a fractal code determine whether or not it converges to a stable image. Methods have been proposed that predict the convergence of a fractal coding process, but to a significant degree they were dependent on the type of fractal encoding scheme used. In this paper we describe how the factors relating to convergence can be calculated for a general class of codes consisting of affine transformations. We demonstrate how it can be implemented and exploited during encoding. Consequently, this new understanding allows us to calculate and control the two factors that reflect convergence, the contractivity and eventual contractivity factors. The main theorems, and the experiments carried out to illustrate their implementation are presented here.

An efficient method is developed to reconstruct broken handwritten digits. The extracted structural points and smoothed skeletons are used to describe the structure of a broken handwritten digit. The broken points of the digits are preselected, based on the minimum distance between the "end" points of two neighboring segments. The correction rules of the preselected broken points are designed, based on an analysis of the structure and skeleton of the digits, and digit reconstruction is made based on the found broken points. Experimental results are given to show the effectiveness of the method.

Presents a fractal transformation based technique for signature verification. In this method, the fractal code of the reference signature locus is applied to a test signature locus and a sequence of fractal decoded test loci is obtained. The successive distances between the decoded test loci transforming towards the fractal attractor are compared to those obtained from the genuine reference signatures. The verification decision is derived from the rate of change of these distances. The experiment shows that the current implementation of the fractal transformation based verification algorithm is effective for random forgery detection.

Presents results of the face verification contest that was organized in conjunction with International Conference on Pattern Recognition 2000. Participants had to use identical data sets from a large, publicly available multimodal database XM2VTSDB. Training and evaluation was carried out according to an a priori known protocol. Verification results of all tested algorithms have been collected and made public on the XM2VTSDB website, facilitating large scale experiments on classifier combination and fusion. Tested methods included, among others, representatives of the most common approaches to face verification -elastic graph matching, Fisher's linear discriminant and support vector machines.

This paper presents a novel algorithm for layout analysis of document images. A major component of this algorithm is the independent segmentation algorithm that identifies text and graphics regions. The segmentation algorithm first locates document patterns and then performs classification using run-length characteristics, spread analysis and adjacency relations. A key feature of the layout analysis algorithm is soft ordering which provides a means of ordering regions in a more logical way, and allows for some overlapping between separate regions. This is very useful for processing documents that are slightly skewed or irregular in layout. The algorithm has been tested on many different documents, and can successfully recognise single and multicolumn documents, even when the column format varies several times on one page. Furthermore, it can process documents with text tightly wrapped around graphics and documents that are slightly skewed.

Fractal image coding has recently been used to perform object recognition, in particular human face recognition. It was shown that the transformations resulting from fractal image coding has invariant properties that can be exploited for recognition. Furthermore, the contractivity factor of a fractal code, which can be used to determine convergence using one code iteration, has a direct effect on the recognition rate. This paper investigates how this rate is affected by the eventual contractivity factor, which is an indicator of guaranteed convergence after more than one iteration of the fractal code. We demonstrate this by ensuring eventual convergence while permitting the contractivity factor to possess values larger than one the recognition rates can be improved. Experiments were performed on the ORL face database and an improved error rate of 1.1% was obtained. We also present a novel method for calculating the eventual contractivity factor for a general class of fractal codes.

A document processing system combining image segmentation with content-based document compression is proposed in the paper. Firstly, a grayscale document image is divided into small blocks and analysed. Then, a modified logical thresholding method based on, local structure analysis and the adaptive logical level technique is used to transform the grayscale document into a binary image. We extract all patterns from the binary document and use a multistage matching method to extract representative patterns. A decomposition method is used to deal with relatively large patterns. Finally, high ratio compression is achieved by coding the relative positions of symbols, extracted representative patterns and other decomposed patterns using the adaptive arithmetic coder anal Q-Coder respectively.

A new method for line image vectorization based on a partition-and-merge technique is presented. A shape is first partitioned into a set of non-overlapping triangles by the constrained Delaunay triangulation. Then the initial partitioning is refined to produce a more accurate result. The vector form of the shape is represented by its skeleton which can be obtained from the skeletons of the parts of the shape. The skeleton thus obtained conforms to human perceptions of the particular shape and can be used to reconstruct the shape.

An intelligent and attractable active contour model for boundary extraction is presented in this paper. The proposed model is capable of driving any initial guess in the area of the evolving estimate towards the desired boundary, working against a constant image background, overcoming spurious edge-points and fitting into the object without any overrun. It is also capable of extracting both concave and convex boundaries while still being capable of bearing subjective boundaries with help of a synthetic convergent criterion and an adaptable interpolation scheme. Using additional two control parameters, it is possible to control the convergent properties of the new model, which provides a high degree of flexibility and adaptability. This robust model has been applied to real images with encouraging results.

In this paper, we propose a new method for computerized human face recognition using fractal transformations. We show that by utilizing the intrinsic properties of block-wise self-similar transformations in fractal image coding we can use it to perform face recognition. The contractivity factor and the encoding scheme of the fractal encoder are shown to affect recognition rates. Using this method, an average error rate of 1.75% was obtained on the ORL face database.

A new method for extracting the boundary of the region of interest (ROI) from motion affected magnetic resonance images (MRI) is presented. An image pre-processing stage is included to suppress prominent ghost artifacts and excessive blurring in the background, in order to facilitate the contour extraction algorithm. The pre-processing stage consists of a novel fuzzy model, incorporating a technique of hierarchical view by view image reconstruction. The contour extraction is performed using an intelligent, attractable active contour model (snakes), which is capable of driving any initial guess in the area of the evolving estimate towards the desired contour, and fitting in to the object without any overrun. The proposed method has been applied to spin echo MRI images affected by rotational motion, producing good results.

Snakes are an effective approach to image segmentation. However, self-looping is a very common problem that makes snakes fail to work well in certain circumstances. In order to achieve robust segmentation, this paper introduces the loop-free snakes based on an attractable active contour model that overcomes several problems of conventional snake model whale retaining all properties associated with it. The proposed method can quickly and eficiently remove all loops during the evolution of snake deformation and can be less sensitive to its parameter setting and flow into more complicated contours such as long tube shapes, sharp corners, deep concave/convex shapes. Hence the new method extends the topologic flexibility and adaptability of snakes. Experiments have been conducted to segment real images with encouraging results.

A dynamic mapping is presented, which is defined by a set of iterative equations. It is shown that the mapping maps the domain space which is composed of all m-dimensional probabilistic vectors to a space which is composed of the m basic unit vectors of the m-dimensional Euclidean space and m(m-1)/spl middot//spl middot//spl middot/(m-j+1)/j! m-dimensional probabilistic vectors in which some components of each probabilistic vector are zero and the remainder are identical. Thus, the dynamic mapping maps the domain space with an infinite number of states to the mapping space which has a finite number of states. The proposed mapping provides an effective classification or cluster scheme when the features of the considered object or data are described by a probabilistic vector.

In this paper, a new fuzzy classifier is developed in which a simple relation between probabilistic vectors and fuzzy sets is derived and the probabilistic relaxation scheme is employed. The fuzzy classifier consists of two stages. Firstly, the fuzzy sets are separated into several groups in terms of the relation between probabilistic vectors and fuzzy sets. Secondly, each group of fuzzy sets is further classified into different subgroups by the probabilistic relaxation scheme. Numerical experiments to verify the effectiveness of the proposed method are carried out. The results show that the method is simple and works well.

This paper presents a fuzzy edge preserving interpolation method for digital images to reduce the effect of jaggedness and blurring artifacts along the high contrast edges. A high subjective performance is achieved by combining two techniques, a region segmentation method and a fuzzy inference method based on the Bayesian structure.

In this paper, a new method is proposed to predict the trend of a stock price based on pattern analysis. The probabilistic relaxation algorithm is employed to classify the probability vectors of the patterns related to a considered stock price. A series of experiments have been carried out on the real stock price to verify the effectiveness of the proposed method.

The adaptive-subspace self-organizing map (ASSOM) proposed by Kohonen is a recent development in self-organizing map (SOM) computation. In this paper, we propose a method to realize ASSOM using a neural learning algorithm in nonlinear autoencoder networks. Our method has the advantage of numerical stability. We have applied our ASSOM model to build a modular classification system for handwritten digit recognition. Ten ASSOM modules are used to capture different features in the ten classes of digits. When a test digit is presented to all the modules, each module provides a reconstructed pattern and the system outputs a class label by comparing the ten reconstruction errors. Our experiments show promising results. For relatively small size modules, the classification accuracy reaches 99.3% on the training set and over 97% on the testing set.

Most handwritten Chinese character recognition systems suffer from the variations in geometrical features for different writing styles. The stroke structures of different styles have proved to be more consistent than geometrical features. In an on-line recognition system, the stroke structure can be obtained according to the sequences of writing via a pen-based input device such as a tablet. But in an off-line recognition system, the input characters are scanned optically and saved as raster images, so the stroke structure information is not available. In this paper, we propose a method to extract strokes from an off-line handwritten Chinese character. We have developed four new techniques: 1) a new thinning algorithm based on Euclidean distance transformation and gradient oriented tracing, 2) a new line approximation method based on curvature segmentation, 3) artifact removal strategies based on geometrical analysis, and 4) stroke segmentation rules based on splitting, merging and directional analysis. Using these techniques, we can extract and trace the strokes in an off-line handwritten Chinese character accurately and efficiently.

A face recognition system is described which employs a fuzzy information fusion technique to increase the overall recognition rate. The face images are searched for locating head area and face boundary. The eyes, and mouth are detected using rigid and deformable templates. Assuming a 3D head model, the face rotations are estimated, which allows for compensating rotated facial features back to a front, upright view. Each facial feature forms a source of information for classification. Based on a correlation technique using eye-forehead, mouth and nose windows, three classifiers are established. The output of each classifier is taken as a partial evidence in classification. The importance of each source is measured using a fuzzy density measure and the final classification is achieved using a fuzzy evidence aggregation method. The performance of the system is evaluated using a combined match score.

The convergence speed of snakes is increased with the switching step size of local minima search and they are incorporated with the local gradient information in the wavelet domain to provide extra speed of convergence to their final contours. This is important in model based compression applications using wavelet decomposition. The directional gradient information is extracted from the subbands provided for the three directions at each stage of the decomposition. The advantages of this approach include, increased speed of convergence due to the switching step size of search and due to the decreased number of pixels to be searched and decreased burden of calculations for pre-processing due to the decreased size of subbands and due to the decreased dimension of the Gaussian filters in the pre-processing stage.

For connected handwritten characters, strokes are merged together in the touching area. Artifacts are often present in the separated characters and therefore the recognition performance will deteriorate. The paper describes an approach to find a partitioning path which can reduce the artifacts significantly.

This paper describes a multi-stage compression scheme of human face images. Snakes are employed in localisation of facial features and biorthogonal spline filters are used for the decomposition of segmented and normalised face images. Wavelet coefficients are vector quantized in different number of cascaded stages depending on their contribution to the subjective quality of the image.

If it is impossible to capture all the image in one scan with the available equipment, a montage can be made from separately scanned pieces. We describe an automatic mosaicing process for document images. The image shifts are found by a correlation technique, using an image pyramid and sequential similarity to reduce computation time. Image placement and overlap is used to reject incorrect solutions. We present results for binarised document images with data captured using a digital camera.

This paper describes a modular classification system for handwritten digit recognition based on the elastic net model. We use ten separate elastic nets to capture different features in the ten classes of handwritten digits and represent an input sample from the activations in each net by population decoding. Compared with traditional neural networks based discriminant classifiers, our scheme features fast training and high recognition accuracy.

This paper propose a vector quantization (VQ) technique to solve the problem of handwritten signature verification. A neural 'gas' model is trained to establish a reference set for each registered person with handwritten signature samples. Then a test sample is compared with all the prototypes in the reference set and the system outputs the label of the writer of the word. Several different feature extraction methods are compared and good results have been obtained by the VQ technique.

A novel stroke extraction method based on a "selective searching" technique is proposed in this paper. A tree structure is constructed in the vicinity of an intersection of strokes. The correct path of a stroke at the intersection can be identified by comparing the travelling cost along each candidate path of the tree. Experimental results show that the method is effective and reliable.

We propose an analytic-to-holistic approach which can identify faces at different perspective variations. The database for the test consists of 40 frontal-view faces. The first step is to locate 15 feature points on a face. A head model is proposed, and the rotation of the face can be estimated using geometrical measurements. The positions of the feature points are adjusted so that their corresponding positions for the frontal view are approximated. These feature points are then compared with the feature points of the faces in a database using a similarity transform. In the second step, we set up windows for the eyes, nose, and mouth. These feature windows are compared with those in the database by correlation. Results show that this approach can achieve a similar level of performance from different viewing directions of a face. Under different perspective variations, the overall recognition rates are over 84 percent and 96 percent for the first and the first three likely matched faces, respectively.

An improved algorithm for planar rotational motion artifact suppression in standard two-dimensional Fourier transform magnetic resonance images is presented. It is shown that interpolation of acquired view data on the uncorrupted k-space create data overlap and void regions. We present a method of managing overlap data regions, using weighted averaging of redundant data. The weights are assigned according to a priority ranking based on the minimum distance between the data set and the k-space grid points. An iterative estimation technique for filling the data void regions, using projections onto convex sets (POCS), is also described. The method has been successfully tested using computer simulations.

An automatic algorithm to extract mouth boundaries in human face images is proposed. The algorithm is based on a hierarchical model adaptation scheme using deformable models. The knowledge about the shape of the object is used to define its initial deformable template. Each mouth boundary curve is initially formed based on three control points whose locations are found through an optimization process using a suitable cost functional. The cost functional captures the essential knowledge about the shape for perceptual organization. Two control points are the mouth corners, which are used as the initial location of the mouth after an approximate mouth window is found based on locating the head boundary. The model is hierarchically improved in the second stage of the algorithm. Each boundary curve is finely tuned using more control points. An old model is adaptively replaced by a new model only if a secondary cost is further reduced. The results show that the model adaptation technique satisfactorily enhances the mouth boundary model in an automated fashion.

After a very promising start, progress in fractal image coding has been relatively slow recently. Most improvements have been concentrating on better adaptive coding algorithms and on search strategies to reduce the encoding time. Very little has been-done to challenge the linear model of the fractal transformations used so far in practical applications. In this paper, we explain why effective nonlinear transformations are not easy to find and propose a model based on conformal mappings in the geometric domain that are a natural extension of the affine model. Our compression results show improvements over the linear model and support the hope that a deeper understanding of the notion of self-similarity would further advance fractal image coding.

In this paper, we present a new approach for detection of brain tumor boundaries in medical images using a Hopfield neural network. The boundary detection problem is formulated as an optimization process that seeks the boundary points to minimize an energy functional based on an active contour model. A modified Hopfield network is constructed to solve the optimization problem. Taking advantage of the collective computational ability and energy convergence capability of the Hopfield network, our method produces the results comparable to those of standard "snakes"-based algorithms, but it requires less computing time. With the parallel processing potential of the Hopfield network, the proposed boundary detection can be implemented for real time processing. Experiments on different magnetic resonance imaging (MRI) data sets show the effectiveness of our approach.

A structural method for describing both printed and handwritten characters is presented in this paper. In this approach, each character is decomposed into primitives based on feature points detection, where a new directional point is introduced and an improved scheme for the bend point detection is proposed. Each primitive is then characterized by a so-called primitive code. A global code is derived from the primitive codes, which is used to describe the topological structure of the character. Each character is thus completely described by the primitive codes and the global code. This method has been successfully applied in handwritten numeral and printed character recognition.

In this paper, the head boundary is first located in a head-and-shoulders image. The approximate positions of the eyes are estimated by means of average anthropometric measures. Corners, the salient features of the eyes, are detected. The corner detection scheme introduced in this paper can provide accurate information about the corners. The shape of the eye is then extracted by a new scheme, based on snakes and on the deformable template. This new representation is in a form similar to snakes, but the energy functional to be minimised is based on some new energy terms and those terms used by the deformable template. A fast algorithm based on the greedy algorithm for active contour modelling is presented for the optimisation step. Experiments show that the execution time required for the whole procedure to extract the eye features is less than one second on a Sun workstation.

An efficient method is presented for automatically locating address block candidates in images of mail pieces. An image is segmented into clusters and the candidate cluster, based on size, position and pixel density consistency is selected for subsequent processing by OCR methods. The success of the method has been demonstrated for several test images.

In this paper, an efficient algorithm is developed for smoothing binary image contours. Using this algorithm we can remove or modify simple spurious points and spurious point groups which have two or more convex or concave points in the direction of a special chain code along a contour. The procedure can be repeated to improve the smoothing quality and the algorithm can be implemented very efficiently.

Presents a technique to produce fuzzy rules based on the ID3 approach and to optimize defuzzification parameters by using a two-layer perceptron. The technique overcomes the difficulties in a conventional syntactic approach to handwritten character recognition, including problems of choosing a starting or reference point, scaling, and learning by machines. The authors' technique provides: a way to produce meaningful and simple fuzzy rules; a method to fuzzify ID3-derived rules to deal with uncertain, noisy, or fuzzy data; and a framework to incorporate fuzzy rules learned from the training data and those extracted from human recognition experience. The authors' experimental results on NIST Special Database 3 show that the technique out-performs the straight forward ID3 approach. Moreover, ID3-derived fuzzy rules can be combined with an optimized nearest neighbor classifier, which uses intensity features only, to achieve a better classification performance than either of the classifiers. The combined classifier achieves a correct classification rate of 98.6% on the test set.

The basic Nearest Neighbor Classifier (NNC) is often inefficient for classification in terms of memory space and computing time needed if all training samples are used as prototypes. These problems can be solved by reducing the number of prototypes using clustering algorithms and optimizing the prototypes using a special neural network model. In this paper, we compare the performance of the multilayer neural network and an Optimized Nearest Neighbor Classifier (ONNC) for handwritten digit recognition applications. We show that an ONNC can have the same recognition performance as an equivalent neural network classifier. The ONNC can be efficiently implemented using prototype and variable ranking, partial summation and distance triangular inequality based strategies. It requires the same memory space as, but less, training time and classification time than the neural network.

Proposes a method for handwritten digit recognition using optimized prototypes generated through learning and transformation. In this method a set of prototypes are obtained from training samples and mapped to a multi-layer neural network for optimization to improve their classification power. The new prototypes are then transformed geometrically to produce a larger set of prototypes for recognition of testing samples. The method has been verified to work well in experimental studies.>

The nearest neighbor rule has previously been shown to be the most reliable method for segmentation of at least a certain range of magnetic resonance images compared with other supervised learning techniques. A nearest neighbor classifier may require long computing time and large memory space if the number of prototypes used is large. The authors present a method for image segmentation using optimized nearest neighbor classifiers. In the method only a very small number of prototypes are generated from training samples using an unsupervised learning method. The prototypes are then optimized using a neural network based on supervised learning. The optimized nearest neighbor classifier is robust in performance for image segmentation and very efficient for practical implementation.>

In this paper, we present a two-layer self-organizing neural network based method for handwritten digit recognition. The network consists of a base layer self-organizing map and a set of corresponding maps in the second layer. The input patterns are partitioned into subspace in the first layer. Patterns in a subspace are led to the second layer and a corresponding map is built according to the first layer performance. In the classification process, each pattern searches for several closest nodes from the base map and then it is classified into a specified class by determining the nearest model of the corresponding maps in the second layer. The new method yielded higher accuracy and faster performance than the ordinary self-organizing neural network.

Kohonen's learning vector quantization (LVQ) is an efficient neural network based technique for pattern recognition. The performance of the method depends on proper selection of the learning parameters. Over-training may cause a degradation in recognition rate of the final classifier. In this paper we introduce constrained learning vector quantization (CLVQ). In this method the updated coefficients in each iteration are accepted only if the recognition performance of the classifier after updating is not decreased for the training samples compared with that before updating, a constraint widely used in many prototype editing procedures to simplify and optimize a nearest neighbor classifier (NNC). An efficient computer algorithm is developed to implement this constraint. The method is verified with experimental results. It is shown that CLVQ outperforms and may even require much less training time than LVQ.