The prevalence of real-world multi-view data makes incomplete multi-view clustering (IMVC) a crucial research. The rapid development of Graph Neural Networks (GNNs) has established them as one of the mainstream approaches for multi-view clustering. Despite significant progress in GNNs-based IMVC, some challenges remain: (1) Most methods rely on the K-Nearest Neighbors (KNN) algorithm to construct static graphs from raw data, which introduces noise and diminishes the robustness of the graph topology. (2) Existing methods typically utilize the Mean Squared Error (MSE) loss between the reconstructed graph and the sparse adjacency graph directly as the graph reconstruction loss, leading to substantial gradient noise during optimization. To address these issues, we propose a novel Dynamic Deep Graph Learning for Incomplete Multi-View Clustering with Masked Graph Reconstruction Loss (DGIMVCM). Firstly, we construct a missing-robust global graph from the raw data. A graph convolutional embedding layer is then designed to extract primary features and refined dynamic view-specific graph structures, leveraging the global graph for imputation of missing views. This process is complemented by graph structure contrastive learning, which identifies consistency among view-specific graph structures. Secondly, a graph self-attention encoder is introduced to extract high-level representations based on the imputed primary features and view-specific graphs, and is optimized with a masked graph reconstruction loss to mitigate gradient noise during optimization. Finally, a clustering module is constructed and optimized through a pseudo-label self-supervised training mechanism. Extensive experiments on multiple datasets validate the effectiveness and superiority of DGIMVCM.

Recently, Vision Transformers (ViTs) have become the state-of-the-art architecture on various computer vision tasks including image classification, object detection and semantic segmentation. However, such success in high-accuracy performance comes at the price of high computational complexity, with typically tens of millions of or even more parameters in a Vision Transformer (ViT) model. Such a large volume of parameters makes it very difficult to deploy ViT models on mobile devices and cumbers their applications. In this paper, we present a novel post-training quantization approach that is able to quantize ViT models to very low bit widths, without the need of re-training. Prior works on post-training quantization for ViTs optimize the quantization of each layer separately thus leading to sub-optimal results. In contrast, we propose a unified learning framework that jointly optimizes the quantization of all layers to directly reduce the overall output error of the network. Moreover, we explore an important property of ViTs, i.e., the additivity property, revealing that the output error caused by the quantization of multiple layers equals the sum of the output error due to the quantization of each layer. Utilizing this property, we present a very efficient algorithm to solve the joint optimization problem with linear time complexity. We performed extensive experiments on the large-scale ImageNet dataset to evaluate the effectiveness of our approach. Empirical results show that our approach improves state-of-the-art noticeably on various ViT models and lowers the bit width from 8-bit to 6-bit without hurting the accuracy. Specifically, at 4 bits, our approach significantly outperforms existing works by 1.72%, 11.49%, 6.15%, and 3.54% on ViT-S, ViT-B, DeiT-S, and DeiT-B, respectively. In the end, we evaluate the performance when deploying our quantized models on hardware. Our approach achieves $1.5\times $ to $1.7\times $ speedups for the inference on NVIDIA A100 GPU.

This article is concerned with stability analysis and observer-based control synthesis for a class of discrete-time switched linear systems with limited statistical information. Instead of commonly studied switching signals such as dwell-time (DT) or Markov chain, a more general class of switching signals, random mode-dependent persistent sojourn-time (RMPST) switching, is investigated. It is composed of fixed parts with no mode switching, random parts without distribution restrictions, and intervals where arbitrary switching is allowed. To tackle the challenges posed by inaccessible modes, unknown transition probabilities, and partially unknown sojourn-time distribution, stability analysis is performed via constructing a Lyapunov function tailored to accommodate the characteristics of RMPST switching. The proposed Lyapunov function is not only mode-dependent but also elapsed-time and quasi-time dependent during fixed-random parts and arbitrary switching intervals, respectively. By means of the new Lyapunov function, an observer-based control approach is introduced for underlying switched systems, and criteria for the existence of observers and controllers are established with the set of admissible switching signals. A novel algorithm is also developed to solve the existence condition by extending the traditional cone complementary linearization (CCL) technique. The effectiveness and applicability of the theoretical results are revealed by a practical space robot manipulator system.

Incomplete multi-view clustering has become one of the important research problems due to the extensive missing multi-view data in the real world. Although the existing methods have made great progress, there are still some problems: 1) most methods cannot effectively mine the information hidden in the missing data; 2) most methods typically divide representation learning and clustering into two separate stages, but this may affect the clustering performance as the clustering results directly depend on the learned representation. To address these problems, we propose a novel incomplete multi-view clustering method with hierarchical information transfer. Firstly, we design the view-specific Graph Convolutional Networks (GCN) to obtain the representation encoding the graph structure, which is then fused into the consensus representation. Secondly, considering that one layer of GCN transfers one-order neighbor node information, the global graph propagation with the consensus representation is proposed to handle the missing data and learn deep representation. Finally, we design a weight-sharing pseudo-classifier with contrastive learning to obtain an end-to-end framework that combines view-specific representation learning, global graph propagation with hierarchical information transfer, and contrastive clustering for joint optimization. Extensive experiments conducted on several commonly-used datasets demonstrate the effectiveness and superiority of our method in comparison with other state-of-the-art approaches.

Incomplete multi-view partial multi-label classification (IMvPMLC), which tackles the combined challenges of incompleteness in both multi-view and multi-label problems, has drawn considerable attention. Existing IMvPMLC methods have made progress but still face several challenges: (i) They mainly focus on the consistency of representations across multiple views but overlook the relationships among instances, leading to suboptimal representations. (ii) They primarily utilize only the available labels for supervised learning, ignoring the missing label distribution and limiting their ability to capture label correlations. In this paper, we propose a novel model named Dual Structure-guided Contrastive Network (DSCN) for IMvPMLC. Specifically, we introduce a similarity-guided instance-level contrastive learning mechanism to achieve multi-view consistent and discriminative representations across instances by leveraging instance structures, while a multi-view attention-based fusion strategy dynamically facilitates the fusion of multi-view representations to derive a robust consensus representation. Then, we design a multi-view shared classifier integrated with a correlation-guided label-level contrastive learning mechanism to enhance predictions by leveraging complementary information across multiple views and capturing label structures, effectively exploiting missing label distribution. Extensive experiments on five benchmark datasets demonstrate that, DSCN yields a more than 13% accuracy, compared with the state-of-the-art approaches. The code and datasets are available at https://anonymous.4open.science/r/DSCN-D471.

In this paper, we propose a post-training pruning framework that jointly optimizes layerwise pruning to minimize model output distortion. Through theoretical and empirical analysis, we discover an important additivity property of output distortion from pruning weights/channels in DNNs. Leveraging this property, we reformulate pruning optimization as a combinatorial problem and solve it with dynamic programming, achieving linear time complexity and making the algorithm very fast on CPUs. Furthermore, we optimize additivity-derived distortions using Hessian-based Taylor approximation to enhance pruning efficiency, accompanied by fine-grained complexity reduction techniques. Our method is evaluated on various DNN architectures, including CNNs, ViTs, and object detectors, and on vision tasks such as image classification on CIFAR-10 and ImageNet, and 3D object detection and various datasets. We achieve SoTA with significant FLOPs reductions without accuracy loss. Specifically, on CIFAR-10, we achieve up to $27.9\times$27.9×, $29.2\times$29.2×, and $14.9\times$14.9× FLOPs reductions on ResNet-32, VGG-16, and DenseNet-121, respectively. On ImageNet, we observe no accuracy loss with $1.69\times$1.69× and $2\times$2× FLOPs reductions on ResNet-50 and DeiT-Base, respectively. For 3D object detection, we achieve $\mathbf {3.89}\times, \mathbf {3.72}\times$3.89×,3.72× FLOPs reductions on CenterPoint and PVRCNN models. These results demonstrate the effectiveness and practicality of our approach for improving model performance through layer-adaptive weight pruning.

Deep neural networks (DNNs) have been widely used in many artificial intelligence (AI) tasks. However, deploying them brings significant challenges due to the huge cost of memory, energy, and computation. To address these challenges, researchers have developed various model compression techniques such as model quantization and model pruning. Recently, there has been a surge in research on compression methods to achieve model efficiency while retaining performance. Furthermore, more and more works focus on customizing the DNN hardware accelerators to better leverage the model compression techniques. In addition to efficiency, preserving security and privacy is critical for deploying DNNs. However, the vast and diverse body of related works can be overwhelming. This inspires us to conduct a comprehensive survey on recent research toward the goal of high-performance, cost-efficient, and safe deployment of DNNs. Our survey first covers the mainstream model compression techniques, such as model quantization, model pruning, knowledge distillation, and optimizations of nonlinear operations. We then introduce recent advances in designing hardware accelerators that can adapt to efficient model compression approaches. In addition, we discuss how homomorphic encryption can be integrated to secure DNN deployment. Finally, we discuss several issues, such as hardware evaluation, generalization, and integration of various compression approaches. Overall, we aim to provide a big picture of efficient DNNs from algorithm to hardware accelerators and security perspectives.

Multi-modality large language models (MLLMs), as represented by GPT-4V, have introduced a paradigm shift for visual perception and understanding tasks, that a variety of abilities can be achieved within one foundation model. While current MLLMs demonstrate primary low-level visual abilities from the identification of low-level visual attributes (e.g., clarity, brightness) to the evaluation on image quality, there's still an imperative to further improve the accuracy of MLLMs to substantially alleviate human burdens. To address this, we collect the first dataset consisting of human natural language feedback on low-level vision. Each feedback offers a comprehensive description of an image's low-level visual attributes, culminating in an overall quality assessment. The constructed Q-Pathway dataset includes 58K detailed human feedbacks on 18,973 multi-sourced images with diverse low-level appearance. To ensure MLLMs can adeptly handle diverse queries, we further propose a GPT-participated transformation to convert these feedbacks into a rich set of 200K instruction-response pairs, termed Q-Instruct. Experimental results indicate that the Q-Instruct consistently elevates various low-level visual capabilities across multiple base models. We anticipate that our datasets can pave the way for a future that foundation models can assist humans on low-level visual tasks.

In this paper, we propose a novel layer-adaptive weight-pruning approach for Deep Neural Networks (DNNs) that addresses the challenge of optimizing the output distortion minimization while adhering to a target pruning ratio constraint. Our approach takes into account the collective influence of all layers to design a layer-adaptive pruning scheme. We discover and utilize a very important additivity property of output distortion caused by pruning weights on multiple layers. This property enables us to formulate the pruning as a combinatorial optimization problem and efficiently solve it through dynamic programming. By decomposing the problem into sub-problems, we achieve linear time complexity, making our optimization algorithm fast and feasible to run on CPUs. Our extensive experiments demonstrate the superiority of our approach over existing methods on the ImageNet and CIFAR-10 datasets. On CIFAR-10, our method achieves remarkable improvements, outperforming others by up to 1.0% for ResNet-32, 0.5% for VGG-16, and 0.7% for DenseNet-121 in terms of top-1 accuracy. On ImageNet, we achieve up to 4.7% and 4.6% higher top-1 accuracy compared to other methods for VGG-16 and ResNet-50, respectively. These results highlight the effectiveness and practicality of our approach for enhancing DNN performance through layer-adaptive weight pruning. Code will be available on https://github.com/Akimoto-Cris/RD_VIT_PRUNE.

A popular track of network compression approach is Quantization aware Training (QAT), which accelerates the forward pass during the neural network training and inference. However, not much prior efforts have been made to quantize and accelerate the backward pass during training, even though that contributes around half of the training time. This can be partly attributed to the fact that errors of low-precision gradients during backward cannot be amortized by the training objective as in the QAT setting. In this work, we propose to solve this problem by incorporating the gradients into the computation graph of the next training iteration via a hypernetwork. Various experiments on CIFAR-10 dataset with different CNN network architectures demonstrate that our hypernetwork-based approach can effectively reduce the negative effect of gradient quantization noise and successfully quantizes the gradients to INT4 with only 0.64 accuracy drop for VGG-16 on CIFAR-10.

Analysis of high dimensional biomedical data such as microarray gene expression data and mass spectrometry images, is crucial to provide better medical services including cancer subtyping, protein homology detection, etc. Clustering is a fundamental cognitive task which aims to group unlabeled data into multiple clusters based on their intrinsic similarities. However, for most clustering methods, including the most widely used K-means algorithm, all features of the high dimensional data are considered equally in relevance, which distorts the performance when clustering high-dimensional data where there exist many redundant variables and correlated variables. In this paper, we aim at addressing the problem of the high dimensional bioinformatics data clustering and propose a new correlation induced clustering, CoIn, to capture complex correlations among high dimensional data and guarantee the correlation consistency within each cluster. We evaluate the proposed method on a high dimensional mass spectrometry dataset of liver cancer tumor to explore the metabolic differences on tissues and discover the intra-tumor heterogeneity (ITH). By comparing the results of baselines and ours, it has been found that our method produces more explainable and understandable results for clinical analysis, which demonstrates the proposed clustering paradigm has the potential with application to knowledge discovery in high dimensional bioinformatics data.

While deep learning methods hitherto have achieved considerable success in medical image segmentation, they are still hampered by two limitations: (i) reliance on large-scale well-labeled datasets, which are difficult to curate due to the expert-driven and time-consuming nature of pixel-level annotations in clinical practices, and (ii) failure to generalize from one domain to another, especially when the target domain is a different modality with severe domain shifts. Recent unsupervised domain adaptation (UDA) techniques leverage abundant labeled source data together with unlabeled target data to reduce the domain gap, but these methods degrade significantly with limited source annotations. In this study, we address this underexplored UDA problem, investigating a challenging but valuable realistic scenario, where the source domain not only exhibits domain shift w.r.t. the target domain but also suffers from label scarcity. In this regard, we propose a novel and generic framework called "Label-Efficient Unsupervised Domain Adaptation" (LE-UDA). In LE-UDA, we construct self-ensembling consistency for knowledge transfer between both domains, as well as a self-ensembling adversarial learning module to achieve better feature alignment for UDA. To assess the effectiveness of our method, we conduct extensive experiments on two different tasks for cross-modality segmentation between MRI and CT images. Experimental results demonstrate that the proposed LE-UDA can efficiently leverage limited source labels to improve cross-domain segmentation performance, outperforming state-of-the-art UDA approaches in the literature.

Multi-label Learning on image data has been widely exploited with deep learning models. However, supervised training on deep CNN models often cannot discover sufficient discriminative features for classification. As a result, numerous self-supervision methods are proposed to learn more robust image representations. However, most self-supervised approaches focus on single-instance single-label data and fall short on more complex images with multiple objects. Therefore, we propose an Object-Aware Self-Supervision (OASS) method to obtain more fine-grained representations for multi-label learning, dynamically generating auxiliary tasks based on object locations. Secondly, the robust representation learned by OASS can be leveraged to efficiently generate Class-Specific Instances (CSI) in a proposal-free fashion to better guide multi-label supervision signal transfer to instances. Extensive experiments on the VOC2012 dataset for multi-label classification demonstrate the effectiveness of the proposed method against the state-of-the-art counterparts.

Image segmentation is one of the most essential biomedical image processing problems for different imaging modalities, including microscopy and X-ray in the Internet-of-Medical-Things (IoMT) domain. However, annotating biomedical images is knowledge-driven, time-consuming, and labor-intensive, making it difficult to obtain abundant labels with limited costs. Active learning strategies come into ease the burden of human annotation, which queries only a subset of training data for annotation. Despite receiving attention, most of active learning methods still require huge computational costs and utilize unlabeled data inefficiently. They also tend to ignore the intermediate knowledge within networks. In this work, we propose a deep active semi-supervised learning framework, DSAL, combining active learning and semi-supervised learning strategies. In DSAL, a new criterion based on deep supervision mechanism is proposed to select informative samples with high uncertainties and low uncertainties for strong labelers and weak labelers respectively. The internal criterion leverages the disagreement of intermediate features within the deep learning network for active sample selection, which subsequently reduces the computational costs. We use the proposed criteria to select samples for strong and weak labelers to produce oracle labels and pseudo labels simultaneously at each active learning iteration in an ensemble learning manner, which can be examined with IoMT Platform. Extensive experiments on multiple medical image datasets demonstrate the superiority of the proposed method over state-of-the-art active learning methods.

Routing protocols for mobile ad-hoc networks have been evaluated extensively through simulation because various network conditions can be easily configured, tested, and replicated across different schemes in simulation than in a real system. Recently, some of these schemes have been implemented in academic, industry and defense testbeds. This gives researchers an opportunity to validate their simulation results with actual implementations. In this paper we report the lessons learned from the implementation of LANMAR (Pei et al. (2000)), a scalable routing protocol that was developed at UCLA as part of large-scale ad hoc network architecture for autonomous unattended agents under ONR support. LANMAR is designed to provide efficient, scalable routing in large ad-hoc wireless networks that exhibit group mobility. In this paper we describe the implementation of this protocol in Linux environments and report on experimental results based on this implementation. The results and lessons from these experiments have enriched our understanding of the LANMAR protocol and its interaction with the other layers and the environment, paving the way to protocol refinements and more efficient implementations.

In this paper, we investigate TCP performance over a multipath routing protocol. Multipath routing can improve the path availability in mobile environment. Thus, it has a great potential to improve TCP performance in ad hoc networks under mobility. Previous research on multipath routing mostly used UDP traffic for performance evaluation. When TCP is used, we find that most times, using multiple paths simultaneously may actually degrade TCP performance. This is partly due to frequent out-of-order packet delivery via different paths. We then test another multipath routing strategy called backup path routing. Under the backup path routing scheme, TCP is able to gain improvements against mobility. We then further study related issues to backup path routing, which can affect TCP performance. Some important discoveries are reported in the paper and simulation results show that by careful selection of the multipath routing strategies, we can improve TCP performance by more than 30% even under very high mobility.

Significant TCP unfairness in ad hoc wireless networks has been reported during the past several years. This unfairness results from the nature of the shared wireless medium and location dependency. If we view a node and its interfering nodes to form a "neighborhood", the aggregate of local queues at these nodes represents the distributed queue for this neighborhood. However, this queue is not a FIFO queue. Flows sharing the queue have different, dynamically changing priorities determined by the topology and traffic patterns. Thus, they get different feedback in terms of packet loss rate and packet delay when congestion occurs. In wired networks, the Randomly Early Detection (RED) scheme was found to improve TCP fairness. In this paper, we show that the RED scheme does not work when running on individual queues in wireless nodes. We then propose a Neighborhood RED (NRED) scheme, which extends the RED concept to the distributed neighborhood queue. Simulation studies confirm that the NRED scheme can improve TCP unfairness substantially in ad hoc networks. Moreover, the NRED scheme acts at the network level, without MAC protocol modifications. This considerably simplifies its deployment.

Emerging wireless ad-hoc networks find their most important applications in untethered, mobile, multihop scenarios where there is no wired infrastructure. Yet, when the wired infrastructure (say, the Internet) is within reach, opportunistic connections to Internet sites may be established across the multihop network to transfer files and update databases. These file transfers use TCP for reliability and congestion control. Many believe that TCP should not be used in ad-hoc network due to it's inability to adapt to the high loss environment, but we believe that TCP will always have a part to play in wireless network in one form or another; studying the behavior of TCP will provide the direction for that adaptation. Recent experiments with ad-hoc, multihop 802.11 networks have exposed serious instabilities when TCP connections span both wired and wireless domains. In particular, some TCP connections capture the wireless channel and drive the throughput on other connections virtually to zero. This is most surprising in view of the fact that connections between 802.11 (single hop) wireless LAN stations and the Internet are well behaved, stable and fair. The problem of the unfairness compounds further when TCP connections have to share the bandwidth with in multihop ad-hoc network. In this paper, we present the issues regarding the wireless transport protocols by experimentally analyzing TCP performance.

IEEE 802.11 MAC mainly relies on two techniques to combat interference: physical carrier sensing and RTS/CTS handshake (also known as "virtual carrier sensing"). Ideally, the RTS/CTS handshake can eliminate most interference. However, the effectiveness of RTS/CTS handshake is based on the assumption that hidden nodes are within transmission range of receivers. In this paper, we prove using analytic models that in ad hoc networks, such an assumption cannot hold due to the fact that power needed for interrupting a packet reception is much lower than that of delivering a packet successfully. Thus, the "virtual carrier sensing" implemented by RTS/CTS handshake cannot prevent all interference. Physical carrier sensing can complement this in some degree. However, since interference happens at receivers, while physical carrier sensing is detecting transmitters (the same problem causing the hidden terminal situation), physical carrier sensing cannot help much, unless a very large carrier sensing range is adopted, which is limited by the antenna sensitivity. We investigate how effective is the RTS/CTS handshake in terms of reducing interference. We show that in some situations, the interference range is much larger than transmission range, where RTS/CTS cannot function well. Then, a simple MAC layer scheme is proposed to solve this problem. Simulation results verify that our scheme can help IEEE 802.11 resolve most interference caused by large interference range.

In ad hoc, multihop wireless networks the routing protocol is key to efficient operation. The design of an ad hoc routing protocol is extremely challenging because of mobility, limited power, unpredictable radio channel behavior and constrained bandwidth. As the network grows large, two additional challenges must be faced: increasing node density, and large number of nodes. High density (i.e., a large number of neighbors within radio range) leads to "superfluous" forwarding of broadcast control messages. Large network size leads to large routing tables and high control traffic overhead. The two aspects are related and they both undermine the scalability of routing protocols. In this paper, we address scalability for a specific class of routing protocols, namely, proactive link state routing protocols. Link state protocols are desirable in many applications because of low access delay, ability to include QoS criteria in path selection, support of alternate routes, etc. Yet, these protocols are most affected by density and large scale. In the paper, we propose two techniques to overcome density and large scale, namely passive clustering and landmark routing. We compare via simulation our proposed solutions to other existing scalable schemes.

A mobile ad hoc network (MANET) is usually assumed to be homogeneous, where each mobile node shares the same radio capacity. However, a homogeneous ad hoc network suffers from poor scalability. Recent research has demonstrated its performance bottleneck both theoretically and through simulation experiments and testbed measurement Building a physically hierarchical ad hoc network is a very promising way to achieve good scalability. In this paper, we present a design methodology to build a hierarchical large-scale ad hoc network using different types of radio capabilities at different layers. In such a structure, nodes are first dynamically grouped into multihop clusters. Each group elects a cluster-head to be a backbone node (BN). Then higher-level links are established to connect the BN into a backbone network. Following this method recursively, a multilevel hierarchical network can be established. Three critical issues are addressed in this paper. We first analyze the optimal number of BN for a layer in theory. Then, we propose a new stable clustering scheme to deploy the BN. Finally LANMAR routing is extended to operate the physical hierarchy efficiently. Simulation results using GloMoSim show that our proposed schemes achieve good performance.

Abstract Secure communication is critical in military environments in which the network infrastructure is vulnerable to various attacks and compromises. A conventional centralized solution breaks down when the security servers are destroyed by the enemies. In this paper we design and evaluate a security framework for multilevel ad hoc wireless networks with unmanned aerial vehicles (UAVs). In battlefields, the framework adapts to the contingent damages on the network infrastructure. Depending on the availability of the network infrastructure, our design is composed of two modes. In infrastructure mode, security services, specifically the authentication services, are implemented on UAVs that feature low overhead and flexible managements. When the UAVs fail or are destroyed, our system seamlessly switches to infrastructureless mode, a backup mechanism that maintains comparable security services among the surviving units. In the infrastructureless mode, the security services are localized to each node's vicinity to comply with the ad hoc communication mechanism in the scenario. We study the instantiation of these two modes and the transitions between them. Our implementation and simulation measurements confirm the effectiveness of our design. Copyright © 2002 John Wiley & Sons, Ltd.

Personal computing on mobile platforms such as laptops and personal digital assistants, rather than in a traditional desktop environment, is becoming increasingly more common. We address the issue of application session transfer for uninterrupted data access across this diverse range of platforms. This work is part of the iMASH project, a multi-year, multi-discipline collaborative effort focused on enabling mobile client platforms and incorporating them into existing legacy networked systems for use by medical practitioners. We have developed a tiered architecture that includes a middleware server layer positioned between existing application servers and multiple clients to make session transfer transparent to the user. Any client application executing our middleware-aware remote code library can save and restore its session by interacting with a middleware server. As a proof of concept, we have implemented the transfer of bookmarks, history, Web cache, and user preferences with the Mozilla open source Web browser, from this effort we have established baseline performance metrics and have found that the overhead is within reasonable bounds of just a few seconds of latency.