While gait recognition based on Large Vision Models (LVMs) benefits from powerful spatial representations, existing paradigms often reduce video sequences to unordered image sets, inherently neglecting the inter-frame temporal causality. To bridge the modal gap between static image perception and dynamic video understanding, we propose the Layer-wise Spatio-Temporal Activation (LTA) framework and the LTA-Gait model. Leveraging the linear complexity of State Space Models (SSMs), our method alleviates the efficiency challenge of introducing temporal modeling into deep LVMs, enabling full-level dense temporal adaptation in a practical manner. Addressing the challenges of recursive noise sensitivity and spatio-temporal granularity misalignment specific to gait video, we design an Anatomy-Aware Temporal Adapter (AATA). By introducing an input-output synergistic dual-gating mechanism, we apply a binary mask-based input constraint to suppress background noise before temporal recursion, while using adaptive soft gating at the output to stabilize feature injection during fine-tuning. Furthermore, we construct a progressive temporal receptive field strategy to better match temporal modeling granularity with the hierarchical spatial features of the LVM. Extensive experiments demonstrate that LTA-Gait achieves State-of-the-Art (SOTA) performance on benchmarks such as CCPG, CCGR, and CASIA-B*. Notably, significant performance gains in challenging clothing-change scenarios validate the core value of robust ordered modeling in extracting intrinsic gait dynamics. All implementation code is fully available at our project repository: https://github.com/zjucgl/LTA-Gait.

Federated Learning (FL) has emerged as a pivotal paradigm for multi-client collaborative learning, primarily due to its inherent capability to safeguard privacy. Nonetheless, the heterogeneity among FL clients, characterized by their disparate resource capabilities and not independent and identical (NonIID) local datasets, presents a significant challenge. Specifically, low-resource clients, such as edge devices, grapple with the inadequacy to accommodate the entire model parameter set for training purposes. To mitigate this issue, preceding research has ventured into devising methodologies that entail extracting sub-models from the overarching global model, tailored to the specific communication, computational, and memory constraints of individual clients. Despite these advancements, prevailing sub-model extraction techniques, which predominantly hinge on pre-established rules, overlook a crucial factor: the impact of NonIID local data on the trajectory of neuron update dynamics. This oversight can amplify discrepancies between the practical local updates inferred by the sub-model and those anticipated via the entire model, thereby undermining overall performance. In this paper, we introduceFedGSE, an innovative Gradient-based Neuron Selection methodology designed explicitly for FL environments. This methodology aims to curate sub-models that significantly reduce discrepancies in local updates, enhancing alignment with the global model's learning trajectory. Central to theFedGSEapproach is a sophisticated algorithm that handpicks critical neurons for sub-model construction. These neurons are identified through their pronounced gradient magnitudes, resulting from the training of the global model on a dataset mirroring the client's data distribution. Consequently, the sub-model's induced local gradient updates closely emulate those derived from directly training the client's data on the full global model, fostering enhanced alignment and performance. Extensive experiments over diverse datasets and tasks demonstrate the superiority ofFedGSEover existing baselines.

Recent research on knowledge distillation has increasingly focused on logit distillation because of its simplicity, effectiveness, and versatility in model compression. In this paper, we introduce Refined Logit Distillation (RLD) to address the limitations of current logit distillation methods. Our approach is motivated by the observation that even high-performing teacher models can make incorrect predictions, creating an exacerbated divergence between the standard distillation loss and the cross-entropy loss, which can undermine the consistency of the student model's learning objectives. Previous attempts to use labels to empirically correct teacher predictions may undermine the class correlations. In contrast, our RLD employs labeling information to dynamically refine teacher logits. In this way, our method can effectively eliminate misleading information from the teacher while preserving crucial class correlations, thus enhancing the value and efficiency of distilled knowledge. Experimental results on CIFAR-100 and ImageNet demonstrate its superiority over existing methods. Our code is available at https://github.com/zju-SWJ/RLD.

High-quality open-source datasets, which necessitate substantial efforts for curation, has become the primary catalyst for the swift progress of deep learning. Concurrently, protecting these datasets is paramount for the well-being of the data owner. Dataset ownership verification emerges as a crucial method in this domain, but existing approaches are often limited to supervised models and cannot be directly extended to increasingly popular unsupervised pre-trained models. In this work, we propose the first dataset ownership verification method tailored specifically for self-supervised pre-trained models by contrastive learning. Its primary objective is to ascertain whether a suspicious black-box backbone has been pre-trained on a specific unlabeled dataset, aiding dataset owners in upholding their rights. The proposed approach is motivated by our empirical insights that when models are trained with the target dataset, the unary and binary instance relationships within the embedding space exhibit significant variations compared to models trained without the target dataset. We validate the efficacy of this approach across multiple contrastive pre-trained models including SimCLR, BYOL, SimSiam, MOCO v3, and DINO. The results demonstrate that our method rejects the null hypothesis with a $p$-value markedly below $0.05$, surpassing all previous methodologies. Our code is available at https://github.com/xieyc99/DOV4CL.

As parallel computing tasks rapidly expand in both complexity and scale, the need for efficient GPU dynamic memory allocation becomes increasingly important. While progress has been made in developing dynamic allocators for substantial applications, their real-world applicability is still limited due to inefficient memory access behaviors. This paper introduces AlignMalloc, a novel memory management system that aligns with the Unified Virtual Memory (UVM) prefetching strategy, significantly enhancing both memory allocation and access performance in large-scale dynamic allocation scenarios. We analyze the fundamental inefficiencies in UVM access and first reveal the mismatch between memory access and UVM prefetching methods. To resolve this issue, AlignMalloc implements a warp-aware memory rearrangement strategy that exploits the regularity of warps to align with the UVM's static prefetching setup. Additionally, AlignMalloc introduces an OR tree-based structure within a host-co-managed framework to further optimize dynamic allocation. Comprehensive experiments demonstrate that AlignMalloc substantially outperforms current state-of-the-art systems, achieving up to$2.7 \times$improvement in dynamic allocation and$2.3 \times$in memory access. Additionally, eight real-world applications with diverse memory access patterns exhibit consistent performance enhancements, with average speedups$1.5 \times$.

Dynamic memory allocation on GPUs, increasingly crucial for applications with dynamic computational patterns, encounters significant challenges due to the complex calculations with intricate branches and substantial memory resources consumed by metadata from massive thread allocations. Despite the current research, there is a lack of a scalable and flexible solution that effectively manages dynamic memory allocation while minimizing memory usage on GPUs. This paper introduces SyncMalloc, a synchronized Host-Device Co-Management system that is specifically designed to adeptly handle dynamic memory allocations of diverse magnitudes. Through the integration of pipelining and producer-consumer mechanisms, SyncMalloc effectively reduces communication overhead and resolves architectural mismatches, further enhancing its capability through synergistic integration with CUDA’s unified memory to facilitate oversubscription. Moreover, SyncMalloc advances slab-based memory management to enhance the efficiency of small allocations, reducing conflict probabilities and overhead in high-activity scenarios. Finally, we present a comprehensive performance evaluation, expanding benchmarks and measurement dimensions to reflect the performance of real-world applications more accurately. The experimental results demonstrate the effectiveness of SyncMalloc in supporting dynamic GPU allocations scaled from 4B to 200GB from multiple perspectives. Our source code is available at https://github.com/jjZhang94/SyncMalloc.

The current pattern mining algorithms focus on discovering either frequent itemsets or high-utility itemsets. The goal of this research is to study the problem of mining frequent-utility itemsets. To solve this problem, two novel algorithms named FUIMTWU-Tree (Frequent-utility Itemset Mining based on TWU-Tree) and FUIMTF-Tree (Frequent-utility Itemset Mining based on TF-Tree) are presented based on the integration of IHUP and HUI-Miner. The TWU-tree and TF-Tree structures are utilised to avoid the unnecessary utility-list construction of itemsets that do not appear in a transaction dataset. The performance of the proposed algorithms is evaluated on various datasets. The results of the experiments demonstrate that FUIMTWU-Tree and FUIMTF-Tree perform efficiently in terms of speed, pruning performance and scalability.

Due to the broad-spectrum and high-efficiency antibacterial activity, antimicrobial peptides (AMPs) and their functions have been studied in the field of drug discovery. Using biological experiments to detect the AMPs and corresponding activities require a high cost, whereas computational technologies do so for much less. Currently, most computational methods solve the identification of AMPs and their activities as two independent tasks, which ignore the relationship between them. Therefore, the combination and sharing of patterns for two tasks is a crucial problem that needs to be addressed. In this study, we propose a deep learning model, called DMAMP, for detecting AMPs and activities simultaneously, which is benefited from multi-task learning. The first stage is to utilize convolutional neural network models and residual blocks to extract the sharing hidden features from two related tasks. The next stage is to use two fully connected layers to learn the distinct information of two tasks. Meanwhile, the original evolutionary features from the peptide sequence are also fed to the predictor of the second task to complement the forgotten information. The experiments on the independent test dataset demonstrate that our method performs better than the single-task model with 4.28% of Matthews Correlation Coefficient (MCC) on the first task, and achieves 0.2627 of an average MCC which is higher than the single-task model and two existing methods for five activities on the second task. To understand whether features derived from the convolutional layers of models capture the differences between target classes, we visualize these high-dimensional features by projecting into 3D space. In addition, we show that our predictor has the ability to identify peptides that achieve activity against Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2). We hope that our proposed method can give new insights into the discovery of novel antiviral peptide drugs.

Determining drug-drug interactions (DDIs) is an important part of pharmacovigilance and has a vital impact on public health. Compared with drug trials, obtaining DDI information from scientific articles is a faster and lower cost but still a highly credible approach. However, current DDI text extraction methods consider the instances generated from articles to be independent and ignore the potential connections between different instances in the same article or sentence. Effective use of external text data could improve prediction accuracy, but existing methods cannot extract key information from external data accurately and reasonably, resulting in low utilization of external data. In this study, we propose a DDI extraction framework, instance position embedding and key external text for DDI (IK-DDI), which adopts instance position embedding and key external text to extract DDI information. The proposed framework integrates the article-level and sentence-level position information of the instances into the model to strengthen the connections between instances generated from the same article or sentence. Moreover, we introduce a comprehensive similarity-matching method that uses string and word sense similarity to improve the matching accuracy between the target drug and external text. Furthermore, the key sentence search method is used to obtain key information from external data. Therefore, IK-DDI can make full use of the connection between instances and the information contained in external text data to improve the efficiency of DDI extraction. Experimental results show that IK-DDI outperforms existing methods on both macro-averaged and micro-averaged metrics, which suggests our method provides complete framework that can be used to extract relationships between biomedical entities and process external text data.

Abstract With continuous advancements in ancient‐book digitization and preservation research, the problems with the stitching of ancient‐book images have become increasingly prominent, as traditional feature‐mapping‐based methods cannot satisfactorily stitch non‐overlapping images. To realize the accurate stitching of the left and right pages of ancient‐book images, this paper proposes a method for ancient‐book image stitching to meet the requirements of their digitization in back‐wrapped binding and other binding forms. First, a dataset of the black text frames from ancient‐book images was established and then used to train a VGG16‐UNet network for the extraction of black text frames. Then, the Douglas–Peucker algorithm was used to fit the black text frames and filter outliers. Finally, a sliding matching algorithm based on the position information of black text frames was proposed for the rectification of misalignments. The results showed that the method achieved a satisfying stitching effect and had good robustness.

3-D human pose estimation or human tracking has always been the focus of research in the human–computer interaction community. As the calibration step of human pose estimation, subject-specific modeling is crucially important to the subsequent pose estimation process. It not only provides a priori knowledge but also clearly defines the tracking target. This article presents a fully automatic subject modeling framework to reconstruct human pose, shape, as well as the body texture in a challenging optimization scenario. By integrating powerful differentiable rendering into the subject-specific modeling pipeline, the proposed method transforms the texture reconstruction problem into analysis by synthesis minimization and solves it efficiently by a gradient-based method. Furthermore, a novel covariance matrix adaptation annealing algorithm is proposed to attack the high-dimensional multimodal optimization problem in an adaptive manner. The domain knowledge of hierarchical human anatomy is seamlessly injected to the annealing optimization process by using a soft covariance matrix mask. All together contributes to the novel algorithm robust to the temptation of local minima. Experiments on the Human3.6 M dataset and the People-Snapshot dataset demonstrate the competitive results to the state of the art both qualitatively and quantitatively.

In the treatment of many diseases, combination drug therapy has been widely used and achieved good clinical efficacy. However, drug-drug interaction (DDI) may occur between multiple drugs and pose a huge threat to the health of patients. Therefore, predicting the presence or absence of DDI among multiple drugs is an important part of pharmacovigilance. Currently, various computational methods for DDI prediction usually use biological information such as molecular structures, targets and enzymes of drugs, or construct heterogeneous networks about drugs, diseases, and genes, so as to obtain abundant information related to drugs. In addition to biological data, pharmacology texts also contain a wealth of information about drug properties, but these texts have not yet been applied to DDI predictions. In this study, we first collect six types of pharmacology texts from DrugBank that can reflect properties of drugs, and propose a novel method named BP-DDI which can combine biological information and pharmacological text to realize DDI event prediction. BP-DDI first extracts biological features (chemical substructure features and target features) from biological data, and then extracts specific types of text features from the collected pharmacology text data. Finally, the biological features are fused with different types of pharmacological text features in order to predict DDI events. Our experiments demonstrate that BP-DDI outperforms existing methods on all three types of prediction tasks. BP-DDI achieves 0.9052 on ACC, and achieves 0.9612 on AUPR.

Summary Nowadays, people use multiple devices to meet the growing requirement for computing. With the application of multicard computing, fault tolerance, load balance, and resource sharing have been the hot issues and the checkpoint/restart (CPR) mechanism is critical in a preemptive system. This article proposes a CPR framework including the automatic compiler (CRAC) to achieve a feasible CPR system, especially for graphics processing unit applications on heterogeneous devices in OpenCL programs. By offering the positions of the CPR in source code, CRAC inserts primitives into programs and invokes the runtime support modules for final results. A comprehensive example and experiments have demonstrated the feasibility and effectiveness of proposed framework.

Checkpoint/restart is an important mechanism to achieve fault tolerance, load balancing and resources sharing in a preemptive system. As Graphics Processing Unit (GPU) becomes quite popular in high performance computing as well as OpenCL programs are portable across various CPUs and GPUs, checkpoint/restart of OpenCL programs on GPUs is in demand. However, due to the intricacy of computation states inside GPUs, there is no effective checkpoint/restart scheme for heterogeneous devices now. This paper proposes a feasible system, CRState, to achieve checkpoint/restart in GPU kernels. With the assistant of a pre-compiler, the primitives are inserted into programs. In run-time, the computation state existing in the underlying hardware is concretized and reconstructed at application level and is ported to heterogeneous devices. Comprehensive experiments have been conducted to demonstrate CRState's feasibility and effectiveness. The experimental results also indicate that CRState has the potential to reschedule resources and balance workload across heterogeneous devices.

Attribute-based encryption (ABE) has been a preferred encryption technology to solve the problems of data protection and access control, especially when the cloud storage is provided by third-party service providers. ABE can put data access under control at each data item level. However, ABE schemes have practical limitations on dynamic attribute revocation. We propose a generic attribute revocation system for ABE with user privacy protection. The attribute revocation ABE (AR-ABE) system can work with any type of ABE scheme to dynamically revoke any number of attributes.

Cloud-based storage is a service model for businesses and individual users that involves paid or free storage resources. This service model enables on-demand storage capacity and management to users anywhere via the Internet. Because most cloud storage is provided by third-party service providers, the trust required for the cloud storage providers and the shared multi-tenant environment present special challenges for data protection and access control. Attribute-based encryption (ABE) not only protects data secrecy, but also has ciphertexts or decryption keys associated with fine-grained access policies that are automatically enforced during the decryption process. This enforcement puts data access under control at each data item level. However, ABE schemes have practical limitations on dynamic user revocation. In this paper, we propose two generic user revocation systems for ABE with user privacy protection, user revocation via ciphertext re-encryption (UR-CRE) and user revocation via cloud storage providers (UR-CSP), which work with any type of ABE scheme to dynamically revoke users.

In some image classification tasks, similarities among different categories are different and the samples are usually misclassified as highly similar categories. To distinguish highly similar categories, more specific features are required so that the classifier can improve the classification performance. In this paper, we propose a novel two-level hierarchical feature learning framework based on the deep convolutional neural network (CNN), which is simple and effective. First, the deep feature extractors of different levels are trained using the transfer learning method that fine-tunes the pre-trained deep CNN model toward the new target dataset. Second, the general feature extracted from all the categories and the specific feature extracted from highly similar categories are fused into a feature vector. Then the final feature representation is fed into a linear classifier. Finally, experiments using the Caltech-256, Oxford Flower-102, and Tasmania Coral Point Count (CPC) datasets demonstrate that the expression ability of the deep features resulting from two-level hierarchical feature learning is powerful. Our proposed method effectively increases the classification accuracy in comparison with flat multiple classification methods.