检索结果-内蒙古大学图书馆

Proceedings of the 41st International Conference on Machine Learning

作者： Wenke Huang Zekun Shi Mang Ye He Li Bo Du National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Wuhan China National Engineering Research Center for Multimedia Software School of Computer Science and Taikang Center for Life and Medical Sciences Wuhan University Wuhan China

Federated learning presents massive potential for privacy-friendly collaboration. However, federated learning is deeply threatened by byzantine attacks, where malicious clients deliberately upload crafted vicious updates. While various robust aggregations have been proposed to defend against such attacks, they are subject to certain assumptions: homogeneous private data and related proxy datasets. To address these limitations, we propose Self-Driven Entropy Aggregation (SDEA), which leverages the random public dataset to conduct Byzantine-robust aggregation in heterogeneous federated learning. For Byzantine attackers, we observe that benign ones typically present more confident (sharper) predictions than evils on the public dataset. Thus, we highlight benign clients by introducing learnable aggregation weight to minimize the instanceprediction entropy of the global model on the random public dataset. Besides, with inherent data heterogeneity, we reveal that it brings heterogeneous sharpness. Specifically, clients are optimized under distinct distribution and thus present fruitful predictive preferences. The learnable aggregation weight blindly allocates high attention to limited ones for sharper predictions, resulting in a biased global model. To alleviate this problem, we encourage the global model to offer diverse predictions via batch-prediction entropy maximization and conduct clustering to equally divide honest weights to accommodate different tendencies. This endows SDEA to detect Byzantine attackers in heterogeneous federated learning. Empirical results demonstrate the effectiveness

关键词：

来源：评论

学校读者我要写书评

暂无评论

Empowering Cross-Device Data Security Verification for IoT Sensor Nodes

Empowering Cross-Device Data Security Verification for IoT S...

引用

Machine Learning and Natural Language Processing (MLNLP), International Conference on

作者： Xin Shi Sheng He Ting Wang Chong Zhang Yang Wang Xiao Zhang School of Computer Science and Software Engineering Southwest Petroleum University Chengdu China School of Computer Science Chengdu University of Information Technology Chengdu China School of Information Engineering Mianyang Teachers̕ College Mianyang China

ISBN: (数字)9798350354973

ISBN: (纸本)9798350354980

In the process of realizing the Internet of Everything through the Internet of Things (IoT), the massive deployment of sensor nodes poses significant security challenges. This paper introduces a novel approach to anomaly detection. The proposed method focuses on generating secure replicas of sensor data, which are compared with real-time data to identify anomalies. This process ensures robust detection across multiple devices and terminals, a key challenge in the realization of the Internet of Everything. The model combines Long Short-Term Memory (LSTM) networks and Variational Autoencoders (VAE), leveraging their strengths in capturing temporal dependencies and learning latent feature representations from sensor data. To enhance detection accuracy in complex environments, an adaptive threshold algorithm is employed, dynamically adjusting thresholds based on real-time data's statistical properties. A crossattention mechanism is also integrated, allowing the model to effectively capture spatio-temporal relationships between sensors. Additionally, we employed a weighted smoothing algorithm to reduce noise while preserving data integrity, ensuring accurate anomaly detection even in noisy environments. Finally, we evaluated our method using the real-world dataset of the Belgian railway bridge KW51. The results indicate that our system can generate sensor data replicas with 99.02% accuracy and detect abnormal data with 99.8% precision. Based on these findings, we believe that this novel approach can effectively perform anomaly detection and provide robust support for IoT security across multiple terminals and devices.

关键词： Adaptation models Accuracy Structural panels Heuristic algorithms Autoencoders Data models Real-time systems Internet of Things Anomaly detection Long short term memory

来源：评论

学校读者我要写书评

暂无评论

Unveiling the Significance of Width Dimension in Bird’s-Eye View Segmentation

Unveiling the Significance of Width Dimension in Bird’s-Eye...

引用

IEEE International Conference on Multimedia and Expo (ICME)

作者： Yiru Wang Qianqian Li Xinyue Wang Qiao Yang Shunli Zhang School of Software Engineering Beijing Jiaotong University Beijing China College of Computer Science and Engineering Shandong University of Science and Technology Shandong China

ISBN: (数字)9798350390155

ISBN: (纸本)9798350390162

Recent advancements in autonomous driving have prominently utilized the bird’s-eye view (BEV) as an intermediary representation of the world. However, with escalating precision requirements, the dimensionality of the BEV representation has expanded, leading to increased complexity. In this paper, we propose a novel real-time BEV segmentation model, which can perform height dimension reduction while view transformation. It also utilizes a novel decoder structure aimed at improving the fusion of overall situations and details. Consequently, we prove that the height dimension for BEV feature extraction and processing tasks is not a necessary component. Experimental results on the nuScenes dataset substantiate that our model surpasses previous approaches in terms of speed and achieves the state-of-the-art performance.

关键词： Dimensionality reduction Accuracy Semantics Feature extraction Real-time systems Decoding Complexity theory

来源：评论

学校读者我要写书评

暂无评论

Temperature-Constrained Computation Time Minimization for Integrated Tasks in UAV-Assisted MEC Networks

Temperature-Constrained Computation Time Minimization for In...

引用

International Conference on Communication Technology (ICCT)

作者： Yuyu Hao Lingyan Bao Hai Li Mingxiong Zhao School of Computer Science and Engineering Macau University of Science and Technology Macau China National Pilot School of Software Yunnan University Kunming China

ISBN: (数字)9798350363760

ISBN: (纸本)9798350363777

Unmanned Aerial Vehicle (UAV)-assisted mobile edge computing (MEC) leverages UAVs’ mobility and cost-effectiveness to handle computation-intensive tasks at the network edge. However, limited energy supply and hardware costs can cause high CPU temperatures due to excessive heat dissipation when processing demanding tasks. To address this, we study a UAV-assisted MEC network where the UAV provides computing services during flight, aiming to minimize users’ average computation time under CPU temperature constraints. We jointly optimize the UAV’s trajectory, resource allocation, and task scheduling, breaking the problem into four subproblems solved iteratively. Simulations confirm that the proposed algorithm meets thermal constraints while reducing average computation time.

关键词： Heating systems Multi-access edge computing Costs Processor scheduling Navigation Autonomous aerial vehicles Minimization Hardware Trajectory Resource management

来源：评论

学校读者我要写书评

暂无评论

Cycle Self-Refinement for Multi-Source Domain Adaptation 38

Cycle Self-Refinement for Multi-Source Domain Adaptation

引用

38th AAAI Conference on Artificial Intelligence, AAAI 2024

作者： Zhou, Chaoyang Wang, Zengmao Du, Bo Luo, Yong School of Computer Science Wuhan University China National Engineering Research Center for Multimedia Software Wuhan University China Institute of Artificial Intelligence Wuhan University China Key Laboratory of Multimedia and Network Communication Engineering Wuhan University China Hubei Luojia Laboratory China

ISBN: (纸本)1577358872

Multi-source domain adaptation (MSDA) aims to transfer knowledge from multiple source domains to the unlabeled target domain. In this paper, we propose a cycle self-refinement domain adaptation method, which progressively attempts to learn the dominant transferable knowledge in each source domain in a cycle manner. Specifically, several source-specific networks and a domain-ensemble network are adopted in the proposed method. The source-specific networks are adopted to provide the dominant transferable knowledge in each source domain for instance-level ensemble on predictions of the samples in target domain. Then these samples with high-confidence ensemble predictions are adopted to refine the domain-ensemble network. Meanwhile, to guide each source-specific network to learn more dominant transferable knowledge, we force the features of the target domain from the domain-ensemble network and the features of each source domain from the corresponding source-specific network to be aligned with their predictions from the corresponding networks. Thus the adaptation ability of source-specific networks and the domain-ensemble network can be improved progressively. Extensive experiments on Office-31, Office-Home and DomainNet show that the proposed method outperforms the state-of-the-art methods for most tasks. Copyright © 2024, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

BiKD-Yolo: A Small Object Detection Algorithm from UAV Perspective Based on BiFormer Attention and Knowledge Distillation

BiKD-Yolo: A Small Object Detection Algorithm from UAV Persp...

引用

International Conference on Fuzzy Systems and Knowledge Discovery (FSKD)

作者： Xiang Hua Zeju Cai Xiangyuan Zhu Ouyang Aijia School of Software Engineering Sun Yat-sen University Guangdong China School of Computer Science and Software Zhaoqing University Guangdong China School of Information Engineering Zunyi Normal University Guizhou China

ISBN: (数字)9798350356328

ISBN: (纸本)9798350356335

Object detection is crucial for the real-time operations of Unmanned Aerial Vehicles (UAVs), particularly in identifying small objects within UAV imagery. While existing computer vision techniques have shown success in UAV applications, detecting small objects remains challenging due to limited information, small size, and low resolution. This study presents the BiKD-YOLO approach, which combines BiFormer attention and knowledge distillation to enhance both the precision and efficiency of small object detection. We employ SAHI (Slice-Assisted Hyper-Inference) cropping to enlarge small objects in images, followed by knowledge distillation for model compression, facilitating knowledge transfer from larger to smaller models for streamlined operation. Through the incorporation of the BiFormer attention mechanism, the detection process is further refined by focusing on critical regions of interest, thereby improving overall accuracy. Experimental Results from tests on the VisDrone dataset show that BiKD-Yolo greatly surpasses the baseline, delivering superior performance in detection accuracy, mAP, and AP metrics. In summary, the BiKD-Yolo method presents a novel and effective solution for improving small object detection in UAV images, offering enhanced accuracy and speed.

关键词： YOLO Measurement Accuracy Refining Streaming media Autonomous aerial vehicles Routing Real-time systems Partitioning algorithms Object recognition

来源：评论

学校读者我要写书评

暂无评论

HSM:Hand Surface Menu Solution for Interaction in VR

HSM:Hand Surface Menu Solution for Interaction in VR

引用

Virtual Reality (ICVR), IEEE International Conference on

作者： Hailong Zhang Lin Li Chang Xue XiaoCai Wei Yin Zhou School of Computer Science and Information Engineering Hefei University of Technology Hefei China School of The Software Engineering Hefei University of Technology Hefei China

ISBN: (数字)9798350364231

ISBN: (纸本)9798350364248

The menu interaction methods in VR, such as floating menus, are still considered unnatural. A solution is proposed in this paper where menus are tightly attached to the user’s palm. Firstly, use UV mapping technology to attach the menu to the palm. Secondly, optimal layout parameters for the hand surface menu were determined through user experiments. Finally, a dataset was built to train the interaction error detection model, aiming to improve interaction success rates under various hand poses in VR. Experimental results demonstrate that compared to other VR menu methods, HSM reduces the average interaction error rate by 30.1% and shortens the average completion time by 12.4% under the optimal hand pose.

关键词： Hands Solid modeling Error analysis Layout Virtual reality Rendering (computer graphics) User experience Surface treatment

来源：评论

学校读者我要写书评

暂无评论

SparseCoder: Identifier-Aware Sparse Transformer for File- Level Code Summarization

SparseCoder: Identifier-Aware Sparse Transformer for File- L...

引用

IEEE International Conference on software Analysis, Evolution and Reengineering (SANER)

作者： Yanlin Wang Yanxian Huang Daya Guo Hongyu Zhang Zibin Zheng School of Software Engineering Sun Yat-sen University China School of Computer Science and Technology Sun Yat-sen University China School of Big Data and Software Engineering Chongqing University China

ISBN: (数字)9798350330663

ISBN: (纸本)9798350330670

Code summarization aims to generate natural language descriptions of source code, facilitating programmers to understand and maintain it rapidly. While previous code summarization efforts have predominantly focused on method-level, this paper studies file-level code summarization, which can assist programmers in understanding and maintaining large source code projects. Unlike method-level code summarization, file-level code summarization typically involves long source code within a single file, which makes it challenging for Transformer-based models to understand the code semantics for the maximum input length of these models is difficult to set to a large number that can handle long code input well, due to the quadratic scaling of computational complexity with the input sequence length. To address this challenge, we propose SparseCoder, an identifier-aware sparse transformer for effectively handling long code sequences. Specifically, the SparseCoder employs a sliding window mechanism for self-attention to model short-term dependencies and leverages the structure message of code to capture long-term dependencies among source code identifiers by introducing two types of sparse attention patterns named global and identifier attention. To evaluate the performance of SparseCoder, we construct a new dataset FILE-CS for file-level code summarization in Python. Experimental results show that our SparseCoder model achieves state-of-the-art performance compared with other pre-trained models, including full self-attention and sparse models. Additionally, our model has low memory overhead and achieves comparable performance with models using full self-attention mechanism. Furthermore, we verify the generality of SparseCoder on other code understanding tasks, i.e., code clone detection and code search, and results show that our model outperforms baseline models in both tasks, demonstrating that our model can generate better code representations for various downstream tasks. Our

关键词： Codes Source coding Computational modeling Semantics Natural languages Cloning Transformers

来源：评论

学校读者我要写书评

暂无评论

Class-wise Adaptive Logits Distillation with Meta-Learning

Class-wise Adaptive Logits Distillation with Meta-Learning

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Xiao Huang Wu Chen Wei Zhou College of Computer and Information Science College of Software Southwest University Chongqing China School of Big Data & Software Engineering Chongqing University Chongqing China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Knowledge distillation has become a crucial technique for transferring intricate knowledge from a teacher model to a smaller student model. While logit-based knowledge distillation has shown promise, existing methods often overlook the efficient distillation of logits. In this paper, we introduce a novel approach called Class-wise Adaptive Logits Distillation (CALD) based on meta-learning. Our method leverages a meta-network to generate class-adaptive weights, delivering both explicit and implicit knowledge adaptively. By training the meta-network to assign higher weights to specific classes crucial for the student model’s learning from the teacher model, our approach enhances the knowledge transfer process. Experimental results on CIFAR-100 and ImageNet datasets demonstrate that CALD surpasses state-of-the-art knowledge distillation methods, achieving enhanced accuracy and efficiency in transferring knowledge from teacher to student models.

关键词： Metalearning Training Adaptation models Accuracy Signal processing Acoustics Speech processing Knowledge transfer

来源：评论

学校读者我要写书评

暂无评论

SparseCoder: Identifier-Aware Sparse Transformer for File-Level Code Summarization

arXiv

引用

arXiv 2024年

作者： Wang, Yanlin Huang, Yanxian Guo, Daya Zhang, Hongyu Zheng, Zibin School of Software Engineering Sun Yat-sen University China School of Computer Science and Technology Sun Yat-sen University China School of Big Data and Software Engineering Chongqing University China

Code summarization aims to generate natural language descriptions of source code, facilitating programmers to understand and maintain it rapidly. While previous code summarization efforts have predominantly focused on method-level, this paper studies file-level code summarization, which can assist programmers in understanding and maintaining large source code projects. Unlike method-level code summarization, file-level code summarization typically involves long source code within a single file, which makes it challenging for Transformer-based models to understand the code semantics for the maximum input length of these models is difficult to set to a large number that can handle long code input well, due to the quadratic scaling of computational complexity with the input sequence length. To address this challenge, we propose SparseCoder, an identifier-aware sparse transformer for effectively handling long code sequences. Specifically, the SparseCoder employs a sliding window mechanism for self-attention to model short-term dependencies and leverages the structure message of code to capture long-term dependencies among source code identifiers by introducing two types of sparse attention patterns named global and identifier attention. To evaluate the performance of SparseCoder, we construct a new dataset FILE-CS for file-level code summarization in Python. Experimental results show that our SparseCoder model achieves state-of-the-art performance compared with other pretrained models, including full self-attention and sparse models. Additionally, our model has low memory overhead and achieves comparable performance with models using full self-attention mechanism. Furthermore, we verify the generality of SparseCoder on other code understanding tasks, i.e., code clone detection and code search, and results show that our model outperforms baseline models in both tasks, demonstrating that our model can generate better code representations for various downstream tasks. Our

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：