High-speed vision tasks have long been a challenge in computer vision. Recently, the spike camera has shown great potential in these tasks due to its high temporal resolution. Unlike traditional cameras, it emits asynchronous spike signals to capture visual information. However, under low-light conditions, spike signals become highly sparse, and the sparse spike stream severely hinders the effectiveness of existing spike-based methods in high-speed scenarios. To address this challenge, we introduce SS2DS, the first deep learning framework that enhances sparse spike streams into dense spike streams. SS2DS first estimates the spike firing frequency within sparse streams. Subsequently, the spike firing frequency is enhanced by a neural network. Finally, SS2DS decodes the enhanced spike stream from the enhanced spike firing frequency sequence. SS2DS can adjust the temporal distribution of sparse spike streams and improve the performance degradation of existing methods in low-light and high-speed scenarios. To evaluate sparse spike stream enhancement, we construct both synthetic and real sparse spike stream datasets. By comparing the reconstruction results, enhanced spike streams achieve an average improvement of +0.78 MA, -18.42 BRISQUE, and -1.42 NIQE over sparse spike streams. Moreover, the enhanced spike streams also benefit other spike-based vision tasks, such as 3D reconstruction (+1.325 dB PSNR, +0.005 SSIM, and -0.01 LPIPS) and superresolution (+0.63 MA, -13.67 BRISQUE, and -1.28 NIQE).

Early diagnosis of Parkinson's disease (PD) is crucial for timely treatment and disease management. Recent studies link PD to impaired facial muscle control, manifesting as "masked face" symptoms, offering a novel diagnostic approach through facial expression analysis. However, data privacy concerns and legal restrictions have resulted in significant "data silos", hindering data sharing and limiting the accuracy and generalizability of existing diagnostic models due to small, localized datasets. To address these challenges, we propose an innovative Tri-Factor Adaptive Federated Learning (TriAFL) framework, designed to collaboratively analyze facial expression data across multiple medical institutions while ensuring robust data privacy protection. TriAFL introduces a comprehensive evaluation mechanism that assesses client contributions across three dimensions: gradient, data, and learning efficiency, effectively addressing Non-IID issues arising from data size variations and heterogeneity. To validate the real-world applicability of our method, we collaborate with a hospital to build the largest known facial expression dataset of PD patients. Furthermore, we explore the integration of local data augmentation strategy to further enhance diagnostic accuracy. Comprehensive experimental results demonstrate TriAFL's superior performance over conventional FL methods in classification task, as well as confirms TriAFL's efficacy in PD diagnosis, delivering a rapid, non-invasive screening tool while driving advancements in AI-powered healthcare.

Federated Learning (FL) is widely used in mobile computing as a communication-efficient distributed machine learning (ML) paradigm; however, it faces challenges such as model convergence to local optima or slow convergence due to the heterogeneity of client data. To mitigate data heterogeneity, the Nesterov Accelerated Gradient (NAG) method demonstrates its effectiveness by predictively updating the gradient to improve system performance. However, the performance of NAG depends heavily on the choice of decay coefficients; larger coefficients have greater acceleration but may lead to an unstable convergence process due to their unreasonable prediction of the descent gradient. To solve the above problems, this paper proposes the first radical heavy ball (RHB) method that combines momentum and NAG. In Stochastic Gradient Descent (SGD), momentum stabilizes the gradient descent process by integrating the historical gradients to update the parameters, and the RHB strategy decouples a single decay coefficient into an NAG component and a momentum component. The RHB introduces a gradient recall after each gradient acceleration by the NAG to strengthen the NAG's perception of the historical gradients, thus stabilizing the gradient descent process. By weighing the historical gradients and the predicted gradient, RHB effectively mitigates the instability of NAG convergence and demonstrates better performance. As a result, the algorithm further mitigates the impact of customer data heterogeneity in FL and can effectively deliver global update information to participants without additional communication costs. We conduct comprehensive experiments in a binary function, single node, and federated model environment to analyze the convergence properties in non-convex loss functions. RHB exhibits better performance and less computational overhead than many existing algorithms.

Hierarchical Federated Learning (HFL) alleviates communication bottlenecks by organizing the system into multiple layers: client, intermediate aggregator, and server. Clients and aggregation layers form groups based on connection patterns, and the methods used for grouping and aggregation directly affect convergence performance. Currently, some studies have proposed grouping algorithms to address the non-independent and identically distributed (non-IID) characteristics of client data to improve performance. However, these methods do not account for network heterogeneity, such as clients using different quantization levels or adaptive quantization strategies to minimize communication overhead. Moreover, most methods rely on heuristics that blindly explore the combinatorial grouping space, incurring substantial computational overhead. In this paper, we conduct a rigorous convergence analysis and frame the dual challenges of heterogeneous data and quantization heterogeneity in HFL as the joint optimization of aggregation weights and grouping strategy. Specifically, we derive the optimal closed-form solution for the aggregation weights and propose an Alternating Optimization Hierarchical Optimal Weights (AO-HOW) algorithm to compute these weights. Building on this result, we propose DyGHFL—Dynamic Exclusion–Reallocation Grouping for Hierarchical Federated Learning, a structu-reaware and efficient greedy algorithm that reorganizes groups by maximizing structured gain and updating weight coefficients according to current system conditions. Experiments on multiple datasets show that DyGHFL consistently outperforms existing baselines, demonstrating its effectiveness in HFL.

Face normalization aims to enhance the robustness and effectiveness of face recognition systems by mitigating intra-personal variations in expressions, poses, occlusions, illuminations, and domains. Existing methods face limitations in handling multiple variations and adapting to cross-domain scenarios. To address these challenges, we propose a novel Unified Multi-Domain Face Normalization Network (UMFN) model, which can process face images with various types of facial variations from different domains, and reconstruct frontal, neutral-expression facial prototypes in the target domain. As an unsupervised domain adaptation model, UMFN facilitates concurrent training on multiple datasets across domains and demonstrates strong prototype reconstruction capabilities. Notably, UMFN serves as a joint prototype and feature learning framework, enabling the simultaneous extraction of domain-agnostic identity features through a decoupling mapping network and a feature domain classifier for adversarial training. Moreover, we design an efficient Heterogeneous Face Recognition (HFR) network that fuses domain-agnostic and identity-discriminative features for HFR, and introduce contrastive learning to enhance identity recognition accuracy. Empirical studies on diverse cross-domain face datasets validate the effectiveness of our proposed method.

Point cloud classifiers have been recently demonstrated to be vulnerable to backdoor attacks. The infected model functions normally on clean data, yet its predictions are errors when triggers are encountered. Currently, the point clean-label backdoor attack (PointCBA) method utilizes feature disentanglement, which is less effective for classes that are not in close proximity to the target class. This paper proposes a novel point cloud backdoor attack approach, named the point clean-label backdoor attack for specific classes (PointCBA-S). PointCBA-S incorporates a strategy named feature entanglement, designed to mitigate the feature similarity between proximate and target classes. This strategy ensures effectiveness across classes distant from the target class. Furthermore, a backdoor spatial optimization mechanism is utilized to create more potent triggers. Experiments indicate that PointCBA-S enhances the average attack success rate (ASR) by 30.8% under different classifiers.

Visual question answering models often rely on language priors in the training set, which affects their generalization ability when dealing with out-of-distribution test data. While existing research has improved model generalization by reducing biased samples, they may compromise model performance. Contrastive learning is a promising solution that uses positive samples with high similarity to a given text to guide the learning process. However, the positive samples are constrained by the size and diversity of existing datasets. We propose CLAP, Contrastive Learning with Answer Perturbation, to enhance model robustness. CLAP generates positive samples by constructing an answer dictionary based on question-image pairs, enabling the model to effectively cluster and differentiate various sample features. Additionally, CLAP introduces answer perturbation to reduce the model’s reliance on potential statistical shortcuts and enhance logical reasoning. Experiments demonstrate that CLAP significantly improves the model’s generalization and robustness while maintaining strong performance and achieving competitive performance. Our code is available at https://github.com/chenyong-cpu/CLAP.

Federated learning is vulnerable to backdoor attacks due to its distributed nature and the inability to access local datasets. Meanwhile, the heterogeneity of distributed data further complicates the detection of such attacks. However, existing defense strategies often overlook the presence of non-stationary objectives and noisy gradients across multiple clients, making it challenging to accurately and efficiently identify malicious participants. To address these challenges, we propose a backdoor defense method for Federated Learning with Adam optimizer and multi-granularity Clustering (FLAC), incorporating both coarse-grained and fine-grained clustering mechanisms to neutralize backdoor attacks. First, the Adam optimizer accelerates the learning process by mitigating the impact of noisy gradients and addressing the non-stationary objectives posed by different clients under attack. Second, a multi-granularity clustering process is considered to differentiate between benign clients and potential attackers. This is followed by an adaptive clipping strategy to further alleviate the influence of malicious attackers. Our theoretical analysis demonstrates the consistent convergence of Adam in a federated backdoor defense environment. Extensive experimental results validate the effectiveness of our defense approach.

Cerebral herniation is a life‐threatening neurological emergency, where timely and accurate prediction is crucial for improving patient prognosis. Due to its rapid imaging advantages, CT becomes the preferred choice for cerebral herniation screening. With the continuous development of artificial intelligence technology in the field of neurological diseases, CT‐based models provide significant support for computer‐aided clinical diagnosis. However, current research on cerebral herniation diagnosis remains limited. Existing methods rely on traditional machine learning or focus solely on midline shift detection, which not only exhibits strong subjectivity but also neglects key structures such as the brainstem and the rich information from sagittal CT images. To address these limitations, this study focuses on mid‐sagittal CT images including the brainstem and combines clinical data to construct a multimodal deep learning framework for cerebral herniation prediction. The model integrates mature and advanced deep learning architectures to extract and fuse features from CT images and clinical text data, employing multiscale convolution and attention mechanisms for diagnostic classification. The model is evaluated on datasets from two centers. Results show that on the internal test set, the model achieves accuracy, sensitivity, specificity, and AUC of 89%, 92%, 88%, and 0.94, respectively; on the external test set, it attains accuracy, sensitivity, specificity, and AUC of 81%, 82%, 80%, and 0.89, respectively, outperforming baseline methods and existing state‐of‐the‐art approaches. Additionally, when compared with radiologists on the internal test set, the model’s performance matches or exceeds the consensus of physicians. We also reveal the model’s focus region through visual analysis, which further deepens the understanding of the model’s prediction process and enhances its interpretability. Experiments demonstrate that the proposed method holds significant potential in assisting cerebral herniation diagnosis.

The accurate mapping of Leaf Area Index (LAI) and Canopy Chlorophyll Content (CCC) is crucial for studying physiological and ecological processes. Leaf Area Index (LAI) is essential for understanding processes such as photosynthesis, transpiration, and energy exchange between the land surface and the atmosphere. CCC serves as an indicator of plant health and productivity, and it is used to assess nitrogen content, which is vital for crop management. We conducted an inversion study for the PROSAIL model using the auto-differentiation (AD) method to retrieve LAI and CCC. The accuracy is assessed on two multispectral/hyperspectral simulated datasets and three measured datasets, covering various types of satellite/airborne image data and in-situ measurements. The RMSE of LAI and CCC is 1.0640 and 74.5518 ug/cm2for multi-spectral simulated datasets and 0.5537 and 39.5267 ug/cm2for hyperspectral simulated datasets under two observation directions. For Barrax agricultural crop datasets, the AD inversion method demonstrates a comparable LAI and an enhanced CCC inversion accuracy when juxtaposed with well-established Simplified Level 2 Product Prototype Processor (SL2P). The feasibility of the AD inversion method is demonstrated in multi-spectral, hyperspectral, and multi-sensor spectral application scenarios. This method has the potential to enhance the retrieval of vegetation biophysical parameters and improve the ecosystem monitoring in the future.

As a powerful tool for simulating bi-directional reflectance and radiative transfer (RT) in complex canopies, the radiosity graphics model (RGM) suffers from a reduced runtime speed or even crashes when facing considerable computation load of the view factor for fine-grained simulation of heterogeneous canopy. In this work, the RGM model at pixel scale (RGMPS model) is proposed with the open accelerator (OpenACC) acceleration techniques and two improved algorithms, which solve the overloaded view factor calculation and enhance scene availability without sacrificing accuracy. Two heterogeneous canopy scenario experiments were used for validation, including a realistic single-tree experiment and a large-scale synthetic heterogeneous canopy experiment. The RGMPS model has increased by nearly 70 times the speed of the original RGM model, demonstrating its capability to model large-scale scenes spanning ten thousand square meters. The${R} ^{2}$between RGM and RGMPS is over 0.94 and the root-mean-square error (RMSE) is below 0.0038. The cross-model validation between the RGMPS model and the discrete anisotropic RT (DART) model achieved a high agreement with${R} ^{2}$as high as 0.98 in the near-infrared (NIR) band. An assessment conducted using airborne multiangle measurements also demonstrated that the accuracy of the proposed solution was deemed satisfactory for bi-directional reflectance factor (BRF) simulation, with RMSEs of 0.0031 and 0.0340 for the red and NIR bands, respectively. Our research contributes to the development of more efficient and accurate BRF simulations for large heterogeneous canopy scenes using the RGM model.

The Moon, with its stable surface reflectance and absence of an atmosphere, is an optimal reference for radiometric calibration of satellite sensors. The Robotic Lunar Observatory (ROLO) model, commonly utilized for this purpose, demonstrates phase dependent imprecision, impacting its predictive accuracy at varying phase angles. This limitation is especially problematic for instruments that cannot select the lunar phase for calibration. This study employs multi-source observational data from ground-based and Himawari-8 sensors to correct the phase dependent imprecision in the ROLO model. Analytical results reveal phase dependent imprecision of 1.89–7.08% across 470–2255 nm in the original ROLO model. After correction, this dependence is reduced to 0.24-0.37%, with a corresponding uncertainty of 3.84%. Utilizing multi-phase-angle lunar observations from FY-2H, the ROLO-corrected model successfully tracked its on-orbit degradation. The results indicate a 6.33–8.64% decline in the radiometric response of FY-2H over 6.5 years, thereby validating the model’s effectiveness in multi-phase-angle calibration. This study presents a robust lunar irradiance reference, particularly suitable for calibrating sensors limited by lunar phase selection, which is critical for long-term monitoring of satellite sensor degradation and enhancing the reliability of quantitative Earth observations.

Hyperspectral remote sensing image feature selection enhances the efficiency of further applications by extracting band information. However, it is challenging to optimize and extract the spectral relationship information of the entire hyperspectral image using traditional methods and solidified spectral representation strategies. As a result, band selection often leads to locally optimal candidate solutions. For instance, when applied to downstream classification tasks, the selected bands typically exhibit issues such as poor information separability, high spectral correlation, and missing information. This paper proposes a new HSIAO_BS framework based on Jeffries-Matusita distance (JM) and an evolutionary algorithm to obtain an excellent subset of bands for hyperspectral remote sensing image feature selection addressing downstream classification tasks. The research problem is modeled as a solution space with effective inter-spectral relationship representation. The HSIAO_BS framework designs an adaptive band encoding mechanism and a feature relationship representation based on JM distance to construct this space. Additionally, the key optimized search method in HSIAO_BS is the improved HSIAO. This evolutionary algorithm combines differential crossover and attenuating mutation strategies to enhance the balance between global exploration and local exploitation capabilities, while also targeting to improve the preference for band selection. The reliability, validity, and stability of the HSIAO_BS framework are verified through a series of performance test experiments conducted on three hyperspectral remote sensing image datasets to support downstream classification tasks.

Wireless sensor networks (WSNs) often utilize virtual backbones (VBs) to optimize routing and reduce energy consumption. The effectiveness of this optimization largely depends on the size of the VB, with smaller VBs offering better performance. In WSNs, VBs are typically modeled as connected dominating sets (CDSs) within unit disk graphs (UDGs). However, existing approximation algorithms for constructing the minimum connected dominating set (MCDS) often introduce redundant dominators, leading to inflated CDSs. To tackle this issue, in this paper, we propose a general CDS optimization algorithm named OP-CDS, designed specifically to minimize redundancies. Theoretical analysis shows that the size of the optimized CDS is bounded by α∙opt+δ-k+1, where α∙opt+δ represents the upper bound of the unoptimized CDS, and k denotes the number of OP-CDS iterations. Additionally, extensive simulations demonstrate that OP-CDS can effectively optimize the CDS generated by state-of-the-art algorithms with minimal time consumption.

For generative learning tasks, there are three crucial criteria for generating samples from the models: quality, coverage/diversity, and sampling speed. Among the existing generative models, Generative adversarial networks (GANs) and diffusion models demonstrate outstanding quality performance while suffering from notable limitations. GANs can generate high-quality results and enable fast sampling, their drawbacks, however, lie in the limited diversity of the generated samples. On the other hand, diffusion models excel at generating high-quality results with a commendable diversity. Yet, its iterative generation process necessitates hundreds to thousands of sampling steps, leading to slow speeds that are impractical for real-time scenarios. To address the aforementioned problem, this paper proposes a novel Consistency-GAN model. In particular, to aid in the training of the GAN, we introduce instance noise, which employs consistency models using only a few steps compared to the conventional diffusion process. Our evaluations on various datasets indicate that our approach significantly accelerates sampling speeds compared to traditional diffusion models, while preserving sample quality and diversity. Furthermore, our approach also has better model coverage than traditional adversarial training methods.

The widespread adoption of deep neural networks (DNNs) is a testament to their profound impact on various domains. However, they are vulnerable to backdoor attacks. Previous defense strategies suffer from requiring additional prior knowledge or performance decreases. To tackle these challenges, we propose a new method to mitigate the impact of backdoor triggers. Specifically, we first devise a simple yet effective detection mechanism based on the region growing algorithm, which enables the identification of triggers within training data without necessitating prior knowledge. Then, we leverage the diffusion model to eliminate the inserted triggers while recovering the data information at the triggers’ locations. Finally, the processed data are fed into the current model for label recovery. Extensive experiments on the CIFAR10, Tiny Imagenet, and GTSRB datasets demonstrate that our method can defend against backdoor attacks effectively and surpasses the state-of-the-art defenses in terms of both main task accuracy (ACC) and backdoor task attack success rate (ASR).

Deep neural networks are vulnerable to adversarial examples, even in black-box scenarios where only the output of the target model is accessible to attackers. This vulnerability poses security concerns, emphasizing the need for early defect detection through adversarial attacks. However, one promising method involves generating adversarial examples on white-box surrogate models to attack the target model based on transferability, which yields a low success rate due to surrogate bias. Directly querying the target model poses efficiency issues and raises another challenge. Furthermore, existing research predominantly focuses on score-based black-box scenarios, assuming query feedback includes scores for all classes. This assumption is impractical in real decision-based scenarios, where only the top-1 class is returned by the target model. To address these challenges, we propose an innovative decision-based attack combining transfer-based and query-based methods, which is robust to surrogate bias. The general idea is to generate adversarial examples utilizing the class-independent skeleton feature on white-box surrogate models, then partially transfer the parameters when querying the target model to adjust, which can solve surrogate model bias. Additionally, we introduce a sampling batch loss to transform top-1 labels into the score distribution, which simplifies our challenging scenarios. Experimental results showcase the efficacy of our method in mitigating surrogate bias, elevating attack success rate and enhancing query efficiency in practical black-box attack scenarios.

MOTIVATION: In drug discovery, it is crucial to assess the drug-target binding affinity (DTA). Although molecular docking is widely used, computational efficiency limits its application in large-scale virtual screening. Deep learning-based methods learn virtual scoring functions from labeled datasets and can quickly predict affinity. However, there are three limitations. First, existing methods only consider the atom-bond graph or one-dimensional sequence representations of compounds, ignoring the information about functional groups (pharmacophores) with specific biological activities. Second, relying on limited labeled datasets fails to learn comprehensive embedding representations of compounds and proteins, resulting in poor generalization performance in complex scenarios. Third, existing feature fusion methods cannot adequately capture contextual interaction information. RESULTS: Therefore, we propose a novel DTA prediction method named HeteroDTA. Specifically, a multi-view compound feature extraction module is constructed to model the atom-bond graph and pharmacophore graph. The residue concat graph and protein sequence are also utilized to model protein structure and function. Moreover, to enhance the generalization capability and reduce the dependence on task-specific labeled data, pre-trained models are utilized to initialize the atomic features of the compounds and the embedding representations of the protein sequence. A context-aware nonlinear feature fusion method is also proposed to learn interaction patterns between compounds and proteins. Experimental results on public benchmark datasets show that HeteroDTA significantly outperforms existing methods. In addition, HeteroDTA shows excellent generalization performance in cold-start experiments and superiority in the representation learning ability of drug-target pairs. Finally, the effectiveness of HeteroDTA is demonstrated in a real-world drug discovery study. AVAILABILITY AND IMPLEMENTATION: The source code and data are available at https://github.com/daydayupzzl/HeteroDTA.

Model quantization has drawn much attention for federated learning (FL) over the Internet of Things (IoT) since it is an effective way to address the critical bottleneck of communication efficiency. State-of-the-art studies have generally assumed homogeneous model quantization, where all clients’ updates are quantized using the same number of bits and aggregated with the same weight at the server. However, in practical IoT scenarios, various IoT devices may apply heterogeneous quantization bits due to their different hardware capabilities, which leads to heterogeneous model quantization accuracy. This article addresses the problem of heterogeneous quantization bit allocation and aggregation for FL. The clients may be allocated with a different number of quantization bits, subject to a total quantization bit constraint. The server may employ different aggregation weights to each IoT device. In particular, we propose FedHBAA—FL with Heterogeneous Bit Allocation and Aggregation, a novel joint quantization bit allocation and server aggregation algorithm. In FedHBAA, we first develop an optimal server aggregation scheme under any given bit allocation among clients. By minimizing the drift term in the convergence rate analysis, a closed-form aggregation weight solution as a function of the allocated bits is obtained for both the strongly convex and nonconvex loss functions. Then, by solving the derived optimal aggregation weights, the optimal bit allocation scheme is obtained. Numerical experiments demonstrate that FedHBAA outperforms the traditional FedAVG algorithm with equal quantization bits.

We integrate the time series simulation capability of the maize model within an extended L-system (ELSYS) using the growth equations from a 4-D maize and a leaf breakpoint model. These models simulate maize growth from emergence to male anthesis, accounting for 3-D architecture during the vegetative season. We employ two methods to achieve time series simulation in ELSYS: directly use the growth equations in the 4-D maize and leaf breakpoint models, and name ELSYS coupling 4-D maize (ELSYS$_{\mathrm {4Dmaize}}$). Alternatively, replace the stem radius-leaf order function with a stem radius-height function and employ linear interpolation to transform the leaf width-length ratio from a constant value to a function that varies with leaf order, thereby simulating a 4-D maize structure, and name the dynamic L-system-based architectural maize (DLAmaize) model. The DLAmaize model is applied to maize canopy reflectance simulations using the discrete anisotropic radiative transfer (DART) model and radiosity-graphics combined method (RGM), along with a comparison with the 1-D scattering by arbitrarily inclined leaves (SAIL) model. The simulated reflectance of maize canopy from ELSYS4Dmaize and DLAmaize differs significantly in the hotspot direction (the absolute value of the relative difference can be up to 67.4%). In addition, comparisons among RT models show that the DART model and RGM simulate close reflectances. The SAIL model yields significant differences (e.g., the absolute value of the relative difference can be up to 41.16% in the nadir direction) owing to its assumption of the homogeneous canopy. DLAmaize enhances remote sensing of dynamic 4-D vegetation canopies modeling and holds promise for remote sensing applications.

The Transformer model, which has significantly advanced natural language processing and computer vision, overcomes the limitations of recurrent neural networks and convolutional neural networks. However, it faces challenges with computational efficiency and memory management due to complex computations and variable-length inputs. Despite research efforts, these issues persist. This paper presents a novel optimization of the Transformer model, following an indepth analysis of its computational graph structure. Firstly, we utilize Deep Computing Unit (DCU) as our hardware platform. Secondly, we optimize element-wise and reduction operators through operator fusion and rewriting. Thirdly, we develop a fine-grained memory management algorithm using a greedy strategy. As a result, the training speed of the Transformer model increases by 1.2x - 1.4x without compromising accuracy.

Due to the progress of large-scale multimodal model pretraining, existing cross-modal retrieval techniques is accurate to align text description to the target image when they show close and clear semantic correspondence. However, in real situations, users only provide ambiguous text queries, making it difficult to retrieve the desired images. To address this issue, we introduce the conversational composed retrieval paradigm, inspired by conversational search which models complex user intent through iterative interaction. This paradigm enhances the model capacity in learning fine-grained correspondences. To train the cross-modal conversational retrieval, we propose the Iterative Refining Retrieval (IRR) framework. It formalizes the reference images and modification texts in each session as a multimodal sequence, which is fed into the generative model to predict the information in the sequence autoregressively, and ultimately predicting the target image feature. In the conversational retrieval paradigm, the model refines the learned correspondences based on the interaction in the later stage of the retrieval session, thus captures fine-grained semantic correspondence to enforce the cross-modal representation. We propose a domain-specific multimodal pretraining method and the full sequence sampling augmentation method to fully utilize the session information. Extensive experiments demonstrate that the iterative refining retrieval method achieves state-of-the-art performance on sessions of varying lengths.

Modern distributed machine learning (ML) paradigms, such as federated learning (FL), utilize data distributed at different clients to train a global model. In such paradigm, local datasets never leave the clients for better privacy protection, and the parameter server (PS) only performs simple aggregation. In practice, however, there is often some amount of data available at the PS, and its computation capability is strong enough to carry out more demanding tasks than simple model aggregation. The focus of this paper is to analyze the model convergence of a new hybrid learning architecture, which leverages the PS dataset and its computation power for collaborative model training with clients. Different from FL where stochastic gradient descent (SGD) is always computed in parallel across clients, the new architecture has both parallel SGD at clients and sequential SGD at PS. We analyze the convergence rate upper bounds of thisaggregate-then-advancedesign for both strongly convex and non-convex loss functions. We show that when the local SGD has an$\mathcal {O}(1/t)$stepsize, the server SGD needs to scale its stepsize to no slower than$\mathcal {O}(1/t^{2})$in order to strictly outperform local SGD with strongly convex loss functions. The theoretical findings are corroborated by numerical experiments, where advantages in terms of both accuracy and convergence speed over clients-only (local SGD and FED AVG) and server-only training are demonstrated.

Nowadays, Federated Learning (FL) over Internet of Medical Things (IoMT) devices has become a current research hotspot. As a new architecture, FL can well protect the data privacy of IoMT devices, but the security of neural network model transmission can not be guaranteed. On the other hand, the sizes of current popular neural network models are usually relatively extensive, and how to deploy them on the IoMT devices has become a challenge. One promising approach to these problems is to reduce the network scale by quantizing the parameters of the neural networks, which can greatly improve the security of data transmission and reduce the transmission cost. In the previous literature, the fixed-point quantizer with stochastic rounding has been shown to have better performance than other quantization methods. However, how to design such quantizer to achieve the minimum square quantization error is still unknown. In addition, how to apply this quantizer in the FL framework also needs investigation. To address these questions, in this paper, we propose FedMSQE - Federated Learning with Minimum Square Quantization Error, that achieves the smallest quantization error for each individual client in the FL setting. Through numerical experiments in both single-node and FL scenarios, we prove that our proposed algorithm can achieve higher accuracy and lower quantization error than other quantization methods.

This paper focuses on an emerging challenging problem called heterogeneous prototype learning (HPL) across face domains-It aims to learn the variation-free target domain prototype for a contaminated input image from the source domain and meanwhile preserve the personal identity. HPL involves two coupled subproblems, i.e., domain transfer and prototype learning. To address the two subproblems in a unified manner, we advocate disentangling the prototype and domain factors in their respected latent feature spaces, and replace the latent source domain features with the target domain ones to generate the heterogeneous prototype. To this end, we propose a disentangled heterogeneous prototype learning framework, dubbed DisHPL, which consists of one encoder-decoder generator and two discriminators. The generator and discriminators play adversarial games such that the generator learns to embed the contaminated image into a prototype feature space only capturing identity information and a domain-specific feature space, as well as generating a realistic-looking heterogeneous prototype. The two discriminators aim to predict personal identities and distinguish between real prototypes versus fake generated prototypes in the source/target domain. Experiments on various heterogeneous face datasets validate the effectiveness of DisHPL.

The task of Visual Question Generation (VQG) aims to generate natural language questions for images. Many methods regard it as a reverse Visual Question Answering (VQA) task. They trained a data-driven generator on VQA datasets, which is hard to obtain questions that can challenge robots and humans. Other methods rely heavily on elaborate but expensive artificial preprocessing to generate. To overcome these limitations, we propose a method to generate inferential questions from the image with noisy captions. Our method first introduces a core scene graph generation module, which can align text features and salient visual features to the initial scene graph. It constructs a special core scene graph with expanded linkage outwards from the high-confidence nodes hop by hop. Next, a question generation module uses the core scene graph as a basis to instantiate the function templates, resulting in questions with varying inferential paths. Experiments show that the visual questions generated by our method are controllable in both content and difficulty, and demonstrate clear inferential properties. In addition, since the salient region, captions, and function templates can be replaced by human-customized ones, our method has strong scalability and potential for more interactive applications. Finally, we use our method to automatically build a new dataset, InVQA, containing about 120k images and 480k question-answer pairs, to facilitate the development of more versatile VQA models.

Vehicular ad hoc networks (VANETs) is one of the most promising approaches for the Intelligent Transportation Systems (ITS). With the rapid increase in the amount of traffic data, deep learning based algorithms have been used extensively in VANETs. The recently proposed federated learning is an attractive candidate for collaborative machine learning where instead of transferring a plethora of data to a centralized server, all clients train their respective local models and upload them to the server for model aggregation. Model quantization is an effective approach to address the communication efficiency issue in federated learning, and yet existing studies largely assume homogeneous quantization for all clients. However, in reality, clients are predominantly heterogeneous, where they support different quantization precision levels. In this work, we propose FedDO – Federated Learning with Double Optimization. Minimizing the drift term in the convergence analysis, which is a weighted sum of squared quantization errors (SQE) over all clients, leads to a double optimization at both clients and server sides. In particular, each client adopts a fully distributed, instantaneous (per learning round) and individualized (per client) quantization scheme that minimizes its own squared quantization error, and the server computes the aggregation weights that minimize the weighted sum of squared quantization errors over all clients. We show via numerical experiments that the minimal-SQE quantizer has a better performance than a widely adopted linear quantizer for federated learning. We also demonstrate the performance advantages of FedDO over the vanilla FedAvg with standard equal weights and linear quantization.

In recent years, the demand for artificial intelligence applications has increased dramatically. Complex models can promote machine learning to achieve excellent results, but computing efficiency has gradually reached a bottleneck. Therefore, more researchers are exploring the improvement of the efficiency of intelligent computing systems. Distributed machine learning can improve the efficiency of model training and inference, but problems such as communication delay and load imbalance between computing nodes still exist. In the multi-GPU distributed computing environment, this paper takes the vision field algorithm VIT (vision transformer) as the optimization object, which has the advantage of convenient parallel training, and proposes several related solutions. Firstly, the parameter server is used as the system logic architecture and in order to reduce the idleness of the computing devices during the training process, the device working status query mechanism is designed to realize load balancing. Secondly, combined with the pre-trained small VIT algorithm model, semi-asynchronous communication method is proposed to reduce the communication overhead of computing devices and accelerate global convergence. The results of this experiment carried out in the existing distributed environment has demonstrated that compared with the existing synchronization method, the computational efficiency has been improved well under the premise of slightly reducing the accuracy.

A prerequisite for solving many agricultural problems is to accurately estimate the area affected by crop disasters and its severity rating. In this letter, we propose a pipeline to segment the drought area and distinguish the severity rating of the maize on RGB images accessed by an unmanned aerial vehicle (UAV) through a semantic segmentation method based on deep learning. First, the ground truth is created through expert evaluation and visual interpretation with the aid of the Normalized Difference Vegetation Index (NDVI). The neural network structure that was used is based on U-Net. Some structural and parameter improvements on U-net were made using SE-ResNeXt-50 as the backbone with the atrous spatial pyramid pooling (ASPP) module. By using RGB images as the input of the neural network for training, the final trained network can work on RGB images captured by a consumer UAV. The experimental results showed that our pipeline achieved an F1-score of 0.9034 and a Jaccard index of 0.8287 on the test set.

Estimation of downward shortwave radiation (DSR) is of great importance in global energy budget and climatic modeling. Although various algorithms have been proposed, effective validation methods are absent for rugged terrains due to the lack of rigorous methodology and reliable field measurements. We propose a two-step validation method for rugged terrains based on computer simulations. The first step is to perform point-to-point validation at local scale. Time-series measurements were applied to evaluate a three-dimensional (3-D) radiative transfer model. The second step is to validate the DSR at pixel-scale. A semiempirical model was built up to interpolate and upscale the DSR. Key terrain parameters were weighted by empirical coefficients retrieved from ground-based observations. The optimum number and locations of ground stations were designed by the 3-D radiative transfer model and Monte Carlo method. Four ground stations were selected to upscale the ground-based observations. Additional three ground stations were set up to validate the interpolated results. The upscaled DSR was finally applied to validate the satellite products provided by MODIS and Himawari-8. The results showed that the modeled and observed DSR exhibited good consistency at point scale with correlation coefficients exceeding 0.995. The average error was around 20 W/m2for the interpolated DSR and 10 W/m2for the upscaled DSR in theory. The accuracies of the satellite products were acceptable at most times, with correlation coefficients exceeding 0.94. From an operational point of view, our method has an advantage of using small amount of ground stations to upscale DSR with relatively high accuracy over rugged terrains.

Communication is widely known as the primary bottleneck of federated learning, and quantization of local model updates before uploading to the parameter server is an effective solution to reduce the communication overhead. However, prior literature always assumes homogeneous quantization for all clients, while in reality, devices are heterogeneous and support different levels of quantization precision. This heterogeneity of quantization poses a new challenge: fine-quantized model updates are more accurate than coarse-quantized ones, and how to optimally aggregate them at the server is an open problem. In this paper, we propose FedHQ – Federated Learning with Heterogeneous Quantization – that allocates different aggregation weights to different clients by minimizing the convergence rate upper bound as a function of the heterogeneous quantization errors of all clients, for both strongly convex and non-convex loss functions. To further accelerate the convergence, the instantaneous quantization error is computed and piggybacked when each client uploads the local model update, and the server dynamically calculates the weight accordingly for the current aggregation. Numerical experiment results demonstrate the performance advantages of FedHQ over both vanilla FedAvg with standard equal weights and a heuristic aggregation scheme, which assigns weights linearly proportional to the clients’ quantization precision.

Quantization of local model updates before uploading to the parameter server is a primary solution to reduce the communication overhead in federated learning. However, prior literature always assumes homogeneous quantization for all clients, while in reality devices are heterogeneous and they support different levels of quantization precision. This heterogeneity of quantization poses a new challenge: fine-quantized model updates are more accurate than coarse-quantized ones, and how to optimally aggregate them at the server is an unsolved problem. In this paper, we propose FEDHQ: Federated Learning with Heterogeneous Quantization. In particular, FEDHQ allocates different weights to clients by minimizing the convergence rate upper bound, which is a function of quantization errors of all clients. We derive the convergence rate of FEDHQ under strongly convex loss functions. To further accelerate the convergence, the instantaneous quantization error is computed and piggybacked when each client uploads the local model update, and the server dynamically calculates the weight accordingly for the current round. Numerical experiments demonstrate the performance advantages of FEDHQ+ over conventional FEDAVG with standard equal weights and a heuristic scheme which assigns weights linearly proportional to the clients’ quantization precision.

Soil-Adjusted Vegetation Index (SAVI) is found to be undesirable to estimate Leaf Area Index (LAI) with heterogeneous canopy structure in low vegetation cover. In this article, three new vegetation indices (VIs), such as Normalized Hotspot-Signature Vegetation Index 2 (NHVI2), Hotspot-Signature Soil-Adjusted Vegetation Index (HSVI), and Hotspot-Signature 2-Band Enhanced Vegetation Index (HEVI2), are proposed for a better quantitative estimation of LAI and soil-noise resistance than with SAVI. To obtain these new indices, the angular index called Normalized Difference between Hotspot and Darkspot (NDHD) is introduced which represents the distribution of foliage in vegetation canopy. The validity of new VIs is statistically verified using simulated data and field measurements. The Discrete Anisotropic Radiative Transfer (DART) model is used to simulate both the homogeneous and heterogeneous canopy for analyzing vegetation isolines behaviors, soil-noise resistance, and LAI estimation. In situ measurements of LAI and bidirectional reflectance factor from the Boreal Ecosystem-Atmosphere Study (BOREAS) are also used to test the robustness of the new VIs for the estimation of LAI. By considering the distribution of the foliage, the accuracy of LAI estimation of SAVI for heterogeneous canopy improved almost 16% using exponential regression analysis. With the improvement of multiangular remote-sensing and Bidirectional Reflectance Distribution Function (BRDF) models in the future, hotspot-signature VIs have the potential to provide a more accurate LAI estimation for heterogeneous canopy in strong soil-noise interference area.

Mare Moscoviense (148°E, 27°N) is one of the few large maria on the lunar farside. In this paper, the China Chang'E-2 Microwave Sounder (CELMS) data were employed to study the microwave thermal emission features of Moscoviense basin. The findings are as follows. (1) The four basaltic units present distinctly different TBperformances at noon and midnight; (2) A new understanding about the basaltic units is made according to the classification results using the maximum likelihood method; (3) The substrate temperature of Moscoviense basin is likely much higher than what we know. The results will be of great significances to understand the mare volcanism on the lunar farside.

In this study, the Chang'E-2 microwave radiometer (CELMS) data are employed to study the thermophysical features of the regolith in the western lunar farside, including Giordano Bruno - the youngest lunar crater of its size. The results are as follows. Firstly, the distribution of the rocks with depth is different in Bruno, King, and Necho craters. Secondly, abundant cold anomalies and one hidden hot anomaly are discovered. Thirdly, at least seven hidden linear structures are discovered, some of which hinting the impacting process of Bruno crater.

Indoor temperature monitoring using wireless sensor networks is critical in many applications. For example, in data centers, it is important to monitor if the indoor temperature is within certain range so that the computers can function with near-optimal performance. In contrast to existing research that often separates the sensor network design and the measured temperature, this paper proposes a cyber-physical design approach to monitor the indoor temperature using wireless sensor networks. The source sensor wakes up and senses the temperature periodically using sleep#x002F;wake duty cycles and sends the data to the destination via multi-hop relaying nodes in an anycast way. Moreover, the period of sleep#x002F;wake duty cycle is dynamically adjusted based on the sensed temperature: when the measured temperature is normal, the sensor nodes wake up infrequently for better energy efficiency; as the sensed temperature approaches a pre-determined threshold, the sensor nodes wake up more frequently to avoid any delayed alarm trigger. The proposed design is implemented using TelosB with TinyOS, and experiments confirm that the cyber-physical system reports the alarm with a very small delay while achieving high long-term energy efficiency.

Security protocols are the key to ensure network security. In the context of the state of the art, so many methods have been developed to analyze the security properties of security protocols, such as Ban logic, theorem proving and model checking etc. This paper used model checking method to formally verify security protocols because of its high degree of automation, briefness and effectiveness. The model checker Spin with sound algorithm design has an extraordinary ability of checking and a good support for LTL. This paper studied the use of Spin on security protocols, and proposed a more effective intruder model to formally verify the security properties of security protocols, such as authentication. The method in this paper decreased the number of model states by a wide margin, and avoided the state space explosion effectively. This paper exampled NSPK protocol and DS protocol, and good experimental results were shown.

The snowmelt runoff model (SRM) is used to simulate streamflow from snowmelt in the Er'dao-Songhua basin of upper Songhuajiang basin from March to August in 2010. The basin is divided into three elevation zones based on SRTM DEM data. MODIS flexible snow cover products (MODISMC) are generated form daily Terra/Aqua products and used as snow cover area input for SRM model. The precipitation and temperature data from climate station are interpolated by Kriging methods to calculate daily average precipitation and temperature for each zone. SRM model is forced by three variables and eight parameters considering the physical and hydrological feature of study area. Results show that the peak of snowmelt runoff comes in the middle of April and the end of May. The Nash-Sutcliffe coefficient of determination (R2) and deviation of the runoff volumes (Dv) is 0.57 and 25.59% respectively. The model errors are mainly caused by ignoring the physical process of snowmelt and lacking enough in-situ materials.

The GPU pass-through technology, under the virtual machine(VM), is always used in CUDA programming. The data will transfer between the VM main memories and GPU memory during the processing of CUDA programs. It is known that the transmission speed of the pinned memory is faster than the pageable when a large amount of data appears. However, the cost of pinned memory will occupy a lot of physical memory and lead to a lack of memory for the VM. So it can also affect the memory usage of other applications in the VM. In this paper, we propose a method to improve the performance of CUDA in virtual environment. In our experiment, the virtual machine manager is chosen as Kernel-based Virtual Machine(KVM), and the GPU pass-through is used for distributing the GPU for the VMs. We also defined a new module as CUDA Memory Management(CMM) which is added into the virtual environment. The result shows that the approach has high efficiency in CUDA program. And the efficiency of data transmission for using the CMM module is more than 16% than the pageable memory only using mode.

Graphical User Interface (GUI) provides a popular and convenient way for the user to freely interact with the systems which makes it widely used in various software applications, it has become an important and indispensable part of today's software. Owing to the characteristics of GUIs different from the traditional software, traditional test techniques and methods cannot satisfy the requirements of GUI testing. Modeling and testing of GUIs-based system is a difficult and challenging work. GUIs-based application is an event-driven application. In GUIs, there exist not only the input events and output events, but also the internal events. In this paper, we identify the input events, output events and internal events and propose an approach to modeling and testing of GUIs-based system using the IOLTS, and input events, output events and internal events are also taken into account. Constraints on events and regular expressions on validation of data are given out. The interactions of GUIs are constructed by the corresponding output events. Finally, tests generation and tests instantiation are given out.

Storage clouds, such as Amazon S3, are being widely used for web services and Internet applications. It has been observed that the delay for retrieving data from and placing data into the clouds is quite random, and exhibits weak correlations between different read/write requests. This inspires us to investigate a key problem: can we reduce the delay by transmitting data replications in parallel or using powerful erasure codes? In this paper, we study the problem of reducing the delay of downloading data from cloud storage systems by leveraging multiple parallel threads, assuming that the data has been encoded and stored in the clouds using fixed rate forward error correction (FEC) codes with parameters (n, k). That is., each file is divided into k equal-sized chunks, which are then expanded into n chunks such that any k chunks out of the n are sufficient to successfully restore the original file. The model can be depicted as a multiple-server queue with arrivals of data retrieving requests and a server corresponding to a thread. However, this is not a typical queueing model because a server can terminate its operation, depending on when other servers complete their service (due to the redundancy that is spread across the threads). Hence, to the best of our knowledge, the analysis of this queueing model remains quite uncharted. Real traces from Amazon S3 show that the time to retrieve a fixed size chunk is random and can be accurately approximated as an i.i.d. exponentially distributed random variable. We show that any work-conserving scheme is delay-optimal when k = 1. When k > 1, we find that a simple greedy scheme, which allocates all available threads to the head of line request, is delay optimal, which appears surprising.

In this paper, we investigate the utility maximization problem for a sensor network with energy replenishment. Each sensor node consumes energy in its battery to generate and deliver data to its destination via multihop communications. Although the battery can be replenished from renewable energy sources, the energy allocation should be carefully designed in order to maximize system performance, especially when the replenishment profile is unknown in advance. In this paper, we address the joint problem of energy allocation and routing to maximize the total system utility, without prior knowledge of the replenishment profile. We first characterize optimal throughput of a single node under general replenishment profile and extend our idea to the multihop network case. After characterizing the optimal network utility with an upper bound, we develop a low-complexity online solution that achieves asymptotic optimality. Focusing on long-term system performance, we can greatly simplify computational complexity while maintaining high performance. We also show that our solution can be approximated by a distributed algorithm using standard optimization techniques. In addition, we show that the required battery size is O(ln(1/ξ)) to constrain the performance of our scheme within ξ-neighborhood of the optimum. Through simulations with replenishment profile traces for solar and wind energy, we numerically evaluate our solution, which outperforms a state-of-the-art scheme that is developed based on the Lyapunov optimization technique.

The smart grid is the new generation of electricity grid that can efficiently utilize new distributed sources of energy (e.g., harvested renewable energy), and allow for dynamic electricity price. In this paper, we investigate the cost minimization problem for an end-user, such as a home, community, or a business, which is equipped with renewable energy devices when electrical appliances allow different levels of delay tolerance. The varying price of electricity presents an opportunity to reduce the electricity bill from an end-user's point of view by leveraging the flexibility to schedule operations of various appliances and HVAC systems. We assume that the end user has an energy storage battery as well as an energy harvesting device so that harvested renewable energy can be stored and later used when the price is high. The energy storage battery can also draw energy from the external grid. The problem we formulate here is to minimize the cost of the energy drawn from the external grid while usage of appliances are subject to individual delay constraints and a long-term average delay constraint. The resulting algorithm requires some future information regarding electricity prices, but it achieves provable performance without requiring future knowledge of either the power demands or the task arrival process. Moreover, we analyze the influence of the assumption that energy can be sold from the battery to the grid. An alternative algorithm is proposed to take advantage of the ability to sell energy. The performance gap between our proposed algorithm and the optimum is shown to diminish as energy selling price approaches the electricity price.

In this paper, we investigate the utility maximization problem for a sensor network with energy replenishment. Each sensor node consumes energy in its battery to generate and deliver data to its destination via multi-hop communications. Although the battery can be replenished from renewable energy sources, the energy allocation should be carefully designed in order to maximize system performance, especially when the replenishment profile is unknown in advance. In this paper, we address the joint problem of energy allocation and routing to maximize the total system utility, without prior knowledge of the replenishment profile. We first characterize optimal throughput of a single node under general replenishment profile, and extend our idea to the multi-hop network case. After characterizing the optimal network utility with an upper bound, we develop a low-complexity online solution that achieves asymptotic optimality. Focusing on long-term system performance, we can greatly simplify computational complexity while maintaining high performance. We also show that our solution can be approximated by a distributed algorithm using standard optimization techniques. Through simulations with replenishment profile traces for solar and wind energy, we numerically evaluate our solution, which outperforms a state-of-the-art scheme that is developed based on the Lyapunov optimization technique.

Network, such as Internet and Intranet, has penetrated into people's daily life. Router is one of the essential equipments which take an important role in the network and form a large and complicated network. However, huge amounts of routers in the network make the network communication and data routing more complex. How to insure the reach ability and correct communication of Internet is a challenge. In this paper, an approach is proposed to formally model and verify the router-based network. Then, we employ a transition system (denotes TS) to model the router-based network, and make use of the Bisimulation-Quotient Algorithms to obtain the bisimulation quotient of the finite transition system, denoted TS/~. It could be easy to verify properties on the system TS/~. Any verification result for TS/~ carries over to TS and this applies to any formula expressed in either LTL, CTL, or CTL*. This approach can facilitate verification since verification problems are particularly space-critical. Finally, some important properties of routing such as routing reach ability, routing path length, are verified.

Driver assistance system is currently an active research and development field for the general goal to provide useful information to drivers in order to reduce the number of injuries and fatalities in traffic. One important component of future systems will be an accurate and reliable positioning, which can be realized by relative measurement, using onboard sensors and maps of the environment. This paper explores the use of dense laser scan data from airborne LiDAR scanning systems for the production of such 3D navigation maps, extracting features from airborne LiDAR data and onboard image data, so as to simultaneous matching and positioning based on overlapping the LiDAR data and onboard images. In order to get the accuracy position, we propose vPOSRIT (vehicle Pose from Orthography and Scaling with Registrations and ITerations) algorithm to determine pose in the establishing correspondence which register 3D architecture features from building models and the real-time images obtained from onboard camera automatically.

In this paper, we investigate the problem of maximizing the throughput over a finite-horizon time period for a sensor network with energy replenishment. The finite-horizon problem is important and challenging because it necessitates optimizing metrics over the short term rather than metrics that are averaged over a long period of time. Unlike the infinite-horizon problem, the fact that inefficiencies cannot be made to vanish to infinitesimally small values, means that the finite-horizon problem requires more delicate control. The finite-horizon throughput optimization problem can be formulated as a convex optimization problem, but turns out to be highly complex. The complexity is brought about by the “time coupling property,” which implies that current decisions can influence future performance. To address this problem, we employ a three-step approach. First, we focus on the throughput maximization problem for a single node with renewable energy assuming that the replenishment rate profile for the entire finite-horizon period is known in advance. An energy allocation scheme that is equivalent to computing a shortest path in a simply-connected space is developed and proven to be optimal. We then relax the assumption that the future replenishment profile is known and develop an online algorithm. The online algorithm guarantees a fraction of the optimal throughput. Motivated by these results, we propose a low-complexity heuristic distributed scheme, called NetOnline, in a rechargeable sensor network. We prove that this heuristic scheme is optimal under homogeneous replenishment profiles. Further, in more general settings, we show via simulations that NetOnline significantly outperforms a state-of-the-art infinite-horizon based scheme, and for certain configurations using data collected from a testbed sensor network, it achieves empirical performance close to optimal.

As Web services are becoming more and more complex, there is an increasing concern about how to guarantee the correctness and safety of Web services composition. This has driven many researchers to study the performance analysis of dynamic atomic service selection, as well as functional verifications. In this paper, we focus on not only modeling the behaviors of atomic service, but also verifying the properties in a quantitative way. First, we apply probabilistic timed model checking to model and verify the behaviors of atomic service by extending interface automata, and propose a technique to formally estimate software performance which exhibits stochastic behaviors with time constrains. Second, the probabilistic timed computation tree logic (PTCTL) formulae are used to express the reliability properties. Third, a failure may occur stochastically when an invocation is triggered through interface operation. We present an internal interaction model, based on which we can dynamically pick out a highest reliable execution sequence for Web services composition. Finally, a case study is demonstrated and experimental results are discussed. In conclusion, our approach provides with an underlying guideline for Web services composition.

As Web applications evolve, their structure may be-come more and more complex. Web frameset is used to organize multiple frames and nested framesets to make the layout of some Web pages more identical and bring the development of Web applications easier, which was wildly used in today's Web applications. How to model and verify the frameset-based Web applications is a challenge. In this paper, special care on Web frameset is paid and an approach to modeling and verifying Web application's navigation with Web frameset is proposed. The Composition semantics of Web Frame-set was give out which can be used to construct complex Web Frameset. Additionally, FSM was employed to describe our models with Web Frameset. Then, we transform FSM model into Kripke structure. Finally, taking advantage of the properties which were generated, we verified our model with Web Framesets. And according to results of verification, we improve our models.

Recent years LiDAR data is widely used for constructing 3D terrain models which provide realistic impressions of the urban environment. This paper presents an automatic method for extracting 3D building model by the fusion of LiDAR data, 2D building outlines and orthoimage. 2D building outlines is generated by classifying the LiDAR data to terrain and off-terrain points, then detecting building edges points through step-structure detector and generalization. 2D building boundaries are added on the DSM (Digital Surface Model) from LiDAR data to generate complex buildings by using CSG with the Boolean operations of union, intersection and differences.

This paper proposes a new model checking-based test generation approach for Web applications. The Kripke structure is reconstructed to model the Web application from the end users' perspective. Test coverage criterion is expressed as trap properties in CTL so that counterexamples can be instantiated to construct test cases. But a counterexample for each trap property is generated will result in too many redundant test cases. So, a test deduction rule and an algorithm based on the greedy heuristic are given to resolve this problem. The test sequences finally generated are those satisfy the coverage criterion and have no redundancy. Throughout the paper, a typical small case study of the WGVS (Web Grade View System) is used to illustrate our approach. This approach presented can help to generate test sequences automatically for Web application and it is a significance complement to the model checking test generation.

As Web applications evolve, their structures become more and more complex. Web browsers may influence on the correctness of the Web applications, and Web browser’s interactions can cause further complications of Web application. Existing navigation models are static ones on the whole. Users’ navigation paths are all determined on stage of model design. Web browser interactions have not been taken into account make them different from practical navigation in Web applications. Moreover, as Web applications evolve and new technologies emerge, adaptive navigation was wildly incorporated in current Web applications. It aggravates the complexity of Web navigations. In this paper, a practical approach to modeling of Web applications and generating tests was pro-posed. And special care is taken on Web browser’s interactions and adaptive navigation during the user’s traversal within hypermedia space. At last, test generation is given out which satisfy the corresponding coverage.

Web applications can only be accessed through dedicated client systems called Web browsers. Most current Web browsers offer many tools or facilities for Web page revisiting, including the back and forward buttons, refresh, favorites, link menu and history lists etc. Users can press the back or forward buttons to negatively influence the behaviors of Web application navigation. Existing navigation models are static ones on the whole. Userspsila navigation paths are all determined on stage of model design. Web browser interactions have not been taken into account make them difference from practical navigation in Web applications. Accordingly, special care is taken on Web browser interactions during the userpsilas traversal within hypermedia space. We give out the concept of safety critical region (SCR) and propose an approach to modeling on-the-fly navigation models. The Kripke structure is employed to describe the on-the-fly navigation models. Coverage criteria of Web browser inter-actions, such as, node coverage, transition coverage triggered by actions, SCR coverage, are exploited to derive the properties of Web browser interactions in CTL. Ultimately, we use SMV, the model checking tool, to verify the on-the-fly navigation models.

Surface emissivity is a measure of the inherent efficiency of the surface to convert heat energy into radiant energy outside the surface. It depends largely on the composition, roughness, and other physical parameters of the surface. The knowledge of surface emissivity permits discrimination and sometimes identification of different types of surfaces. The largest area of mountainous permafrost is on the Tibet plateau across the world, except the polar regions. Cloud free thermal infrared (TIR) data from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Terra satellite in early winter (Beijing time: Oct.20, 2001, 14:46:25) were chosen to retrieve the surface emissivity on the Tibet plateau by employing a spectral method (the emissivity normalization, NOR). The surface emissivity images were classified by an un-supervised classification method, and some segments were merged into a permafrost distribution map, which includes continuous permafrost, isolated 'island' permafrost, seasonal permafrost (around the mountain or along the valley), and snow-ice. The results are rigorously consistent with field survey. And more details of the distribution of permafrost are also possible by this method.

Based on the World-Wide-Web (WWW) and Internet, the virtual geographical environment (VGE) is a virtual and three-dimensional environment shared for multiusers, where multidimensional geo-scientific data can be released, and geo-scientific process or phenomena can be simulated. A prototype of VGE for the Tibet Plateau has been conceived and is being built, under which the evolving processes of the Tibet Plateau can be simulated by a user's interactions. The multisource geo-scientific information including oil-gas seepages, active structures, permafrost distribution, earthquakes, and so on, were integrated successfully to predict oil-gas or gas-hydrates under the permafrost by the application of the prototype. The topography over the Tibet Plateau is visualized three-dimensionally by overlaying color composite image of remote sensing on to a digital elevation model (DEM) according to the scales the users assume. The three-dimensional information and the results simulated can be Web browsed by a Web plug-in.

Land-cover classification represents one of the most fundamental applications of remote sensing, and is widely used to estimate carton stocks and parameters hydrological and biogeochemical models. Several studies reveal that changing the spatial resolution of land-cover maps has important effects on the proportion of a landscape occupied by a particular land cover type. We study the proportions of vegetations based on multi-scale land-cover classifications in the area of Qianyanzhou in the province of Jiangxi in the South China on the base of ground investigations. The viability of coarse spatial resolution data for land-cover classification is evaluated using degraded Landsat Thematic Mapper (TM). The uncertainties of multi-scale land cover classifications are finally analyzed based on the different aggregated TM land-cover maps.