检索结果-内蒙古大学图书馆

LoCo: Low-Bit Communication Adaptor for Large-Scale Model Training

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年第6期47卷 4285-4298页

作者： Xie, Xingyu Lin, Zhijie Toh, Kim-Chuan Zhou, Pan National University of Singapore Department of Mathematics Singapore Singapore Management University School of Computing and Information Systems Singapore Sea Ai Lab Singapore

To efficiently train large-scale models, low-bit gradient communication compresses full-precision gradients on local GPU nodes into low-precision ones for higher gradient synchronization efficiency among GPU nodes. However, it often degrades training quality due to compression information loss. To address this, we propose the Low-bit Communication Adaptor (LoCo), which compensates gradients on local GPU nodes before compression, ensuring efficient synchronization without compromising training quality. Specifically, LoCo designs a moving average of historical compensation errors to stably estimate concurrent compression error and then adopts it to compensate for the concurrent gradient compression, yielding a less lossless compression. This mechanism allows it to be compatible with general optimizers like Adam and sharding strategies like FSDP. Theoretical analysis shows that integrating LoCo into full-precision optimizers like Adam and SGD does not impair their convergence speed on non-convex problems. Experimental results show that across large-scale model training frameworks like Megatron-LM and PyTorch's FSDP, LoCo significantly improves communication efficiency, e.g., improving Adam's training speed by 14% to 40% without performance degradation on large language models like LLAMAs and MoEs. © 1979-2012 IEEE.

关键词： Synchronization

来源：评论

学校读者我要写书评

暂无评论

TeraPRI: Homomorphic Terahertz-Empowered Joint Wireless Power and information Transfer with Privacy-Preserving for 6G-Autonomous Vehicles

引用

IEEE Transactions on Consumer Electronics 2025年

作者： Pauu, Kulaea Taueveeve Wu, Jun Bashir, Ali Kashif Waseda University Graduate School of Information Production and Systems Japan Manchester Metropolitan University Department of Computing and Mathematics United Kingdom

In sixth-generation networks, Terahertz technology enables high-speed, low-latency communication capabilities, while Mobile Edge computing (MEC) enhances remote computation, leveraging Autonomous Vehicles capable of providing computational resources and energy support, particularly in challenging environments where conventional terrestrial infrastructure is absent. However, challenges remain, particularly with the limited energy of end-user nodes and the constrained data capabilities of MEC nodes in THz networks, which require substantial energy to transmit large volumes of data. Efficient coordination of energy supply and information transfer is essential. Additionally, such networks can face signal interference and leakage of privacy-sensitive information, complicating data security and reliable communication. To address these challenges, we propose TeraPRI, a novel framework for homomorphic THz-empowered joint wireless power and information transfer with privacy preservation for 6G Autonomous Vehicles. First, we introduce a dual-mode adaptive resource allocation method that alternates between »harvest-and-then-transmit-mode» (HaT-mode) and »transmit-and-then-harvest-mode» (TaH-mode) based on end-user node demands. Second, we design a selective CKKS homomorphic encryption technique with lightweight thresholding for frequency allocation, enabling the MEC to securely assign unique THz sub-channels to end-user nodes in the low-THz band (0.1-1.0 THz) while enhancing communication security through randomized frequency hopping. Extensive simulations show that TeraPRI significantly improves power transfer, information transfer rates, and privacy over existing methods. © 1975-2011 IEEE.

关键词： Inductive power transmission

来源：评论

学校读者我要写书评

暂无评论

Admitting Ignorance Helps the Video Question Answering Models to Answer

arXiv

引用

arXiv 2025年

作者： Li, Haopeng Drummond, Tom Gong, Mingming Bennamoun, Mohammed Ke, Qiuhong The School of Computing and Information Systems University of Melbourne Australia The School of Mathematics and Statistics University of Melbourne Australia The school of Physics Maths and Computing The University of Western Australia Australia The Department of Data Science & AI Monash University The School of Computing and Information Systems University of Melbourne Australia

Significant progress has been made in the field of video question answering (VideoQA) thanks to deep learning and large-scale pretraining. Despite the presence of sophisticated model structures and powerful video-text foundation models, most existing methods focus solely on maximizing the correlation between answers and video-question pairs during training. We argue that these models often establish shortcuts, resulting in spurious correlations between questions and answers, especially when the alignment between video and text data is suboptimal. To address these spurious correlations, we propose a novel training framework in which the model is compelled to acknowledge its ignorance when presented with an intervened question, rather than making guesses solely based on superficial question-answer correlations. We introduce methodologies for intervening in questions, utilizing techniques such as displacement and perturbation, and design frameworks for the model to admit its lack of knowledge in both multi-choice VideoQA and open-ended settings. In practice, we integrate a state-of-the-art model into our framework to validate its effectiveness. The results clearly demonstrate that our framework can significantly enhance the performance of VideoQA models with minimal structural modifications. Copyright © 2025, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Navigating Ethical Dimensions in the Metaverse: Challenges, Frameworks, and Solutions

引用

IEEE Access 2025年 13卷 79996-80018页

作者： Al-Kfairy, Mousa Alrabaee, Saed Alfandi, Omar Mohamed, Amr Taha Khaddaj, Souheil Zayed University College of Technological Innovation Abu Dhabi United Arab Emirates United Arab Emirates University Information Systems Security Al Ain15551 United Arab Emirates Zayed University College of Humanities and Social Sciences Abu Dhabi United Arab Emirates Kingston University School of Computing Science and Mathematics Kingston Upon Thames United Kingdom

The Metaverse is rapidly evolving into a transformative digital ecosystem, bringing with it unprecedented opportunities and a complex array of ethical challenges. This narrative review, based on an in-depth analysis of 105 full publications, explores the key ethical themes associated with the Metaverse, including privacy and data security, identity and behavior, digital inclusivity, mental and physical health, ethical AI, content moderation, intellectual property, governance, environmental sustainability, harassment, cultural representation, and economic implications. Proposed solutions for these challenges encompass privacy-by-design frameworks, robust identity verification systems, equitable access initiatives, explainable AI, and blockchain-based intellectual property protections. Additionally, the review examines governance and legal initiatives, such as frameworks developed by the World Economic Forum, European Commission, United Nations, and IEEE standards. These efforts aim to enhance transparency, ensure universal accessibility, promote ethical AI, and foster safety within the Metaverse. By synthesizing insights from this extensive body of literature, this review offers a thorough exploration of the ethical dimensions of the Metaverse and presents practical recommendations for fostering responsible and inclusive digital ecosystems. © 2013 IEEE.

关键词： Ethical technology

来源：评论

学校读者我要写书评

暂无评论

A General Framework for Clustering and Distribution Matching with Bandit Feedback

引用

IEEE Transactions on information Theory 2025年第3期71卷 2116-2139页

作者： Yavas, Recep Can Huang, Yuqi Tan, Vincent Y. F. Scarlett, Jonathan CNRS@CREATE DesCartes Program CREATE Tower 138602 Singapore National University of Singapore Department of Computer Science Queenstown119077 Singapore National University of Singapore School of Computing Department of Information Systems and Analytics Queenstown119077 Singapore National University of Singapore Department of Mathematics Department of Electrical and Computer Engineering Queenstown119077 Singapore Institute of Data Science National University of Singapore Department of Computer Science The Department of Mathematics Queenstown119077 Singapore

We develop a general framework for clustering and distribution matching problems with bandit feedback. We consider a K-armed bandit model where some subset of K arms is partitioned into M groups. Within each group, the random variable associated to each arm follows the same distribution on a finite alphabet. At each time step, the decision maker pulls an arm and observes its outcome from the random variable associated to that arm. Subsequent arm pulls depend on the history of arm pulls and their outcomes. The decision maker has no knowledge of the distributions of the arms or the underlying partitions. The task is to devise an online algorithm to learn the underlying partition of arms with the least number of arm pulls on average and with an error probability not exceeding a pre-determined value δ. Several existing problems fall under our general framework, including finding M pairs of arms, odd arm identification, and N-ary clustering of K arms belong to our general framework. We derive a non-asymptotic lower bound on the average number of arm pulls for any online algorithm with an error probability not exceeding δ. Furthermore, we develop a computationally-efficient online algorithm based on the Track-and-Stop method and Frank-Wolfe algorithm, and show that the average number of arm pulls of our algorithm asymptotically matches that of the lower bound. Our refined analysis also uncovers a novel bound on the speed at which the average number of arm pulls of our algorithm converges to the fundamental limit as δ vanishes. © 1963-2012 IEEE.

关键词： Random variables

来源：评论

学校读者我要写书评

暂无评论

AttentionPoolMobileNeXt: An automated construction damage detection model based on a new convolutional neural network and deep feature engineering models

引用

Multimedia Tools and Applications 2025年第4期84卷 1821-1843页

作者： Aydin, Mehmet Barua, Prabal Datta Chadalavada, Sreenivasulu Dogan, Sengul Tuncer, Turker Chakraborty, Subrata Acharya, Rajendra U. Department of Digital Forensics Engineering Technology Faculty Firat University Elazig Turkey University of Southern Queensland Toowoomba Australia School of Engineering University of Southern Queensland Springfield LakesQLD4300 Australia School of Science and Technology Faculty of Science Agriculture Business and Law University of New England ArmidaleNSW2351 Australia Center for Advanced Modelling and Geospatial Information Systems Faculty of Engineering and IT University of Technology Sydney SydneyNSW2007 Australia School of Mathematics Physics and Computing University of Southern Queensland Toowoomba Australia

In 2023, Turkiye faced a series of devastating earthquakes and these earthquakes affected millions of people due to damaged constructions. These earthquakes demonstrated the urgent need for advanced automated damage detection models to help people. This study introduces a novel solution to address this challenge through the AttentionPoolMobileNeXt model, derived from a modified MobileNetV2 architecture. To rigorously evaluate the effectiveness of the model, we meticulously curated a dataset comprising instances of construction damage classified into five distinct classes. Upon applying this dataset to the AttentionPoolMobileNeXt model, we obtained an accuracy of 97%. In this work, we have created a dataset consisting of five distinct damage classes, and achieved 97% test accuracy using our proposed AttentionPoolMobileNeXt model. Additionally, the study extends its impact by introducing the AttentionPoolMobileNeXt-based Deep Feature Engineering (DFE) model, further enhancing the classification performance and interpretability of the system. The presented DFE significantly increased the test classification accuracy from 90.17% to 97%, yielding improvement over the baseline model. AttentionPoolMobileNeXt and its DFE counterpart collectively contribute to advancing the state-of-the-art in automated damage detection, offering valuable insights for disaster response and recovery efforts. © The Author(s) 2024.

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

Interval-Shared information Integration and False-Negative Association Reduction in Multi-Source MiRNA-Disease Association Prediction

引用

IEEE Journal of Biomedical and Health Informatics 2025年 PP卷 PP页

作者： Cui, Qinghang Guo, Honglie Cai, Yueyi Fei, Yu Wang, Shunfang Yunnan University School of Information Science and Engineering Yunnan Kunming650500 China Yunnan University of Finance and Economics School of Statistics and Mathematics Yunnan Kunming650221 China Key Lab of Intelligent Systems and Computing of Yunnan Province Yunnan Kunming650500 China

Numerous studies have demonstrated that microRNAs (miRNAs) play crucial roles in the development and progression of various diseases, making the identification of miRNA-disease associations (MDAs) essential for understanding human disease etiology. While several computational models have been developed to predict MDAs, challenges persist—particularly the limited consideration of information interactions among multi-source similarities and the presence of "false-negative" associations in the original topology. To address these issues, we propose ISFNMDA, a model designed to infer potential MDAs by leveraging multi-view collaborative learning for feature extraction and optimizing association topology through graph structure momentum contrastive learning. Specifically, multi-source similarities of miRNAs and diseases are mapped into a unified feature space via encoders. The Pearson correlation coefficient is employed to derive pairwise constraints between nodes, facilitating information interactions and constructing interval-shared information constraints. Subsequently, an inference graph learner models the representations to generate an inferred graph topology. By maximizing mutual information between the inferred topology and the original "false-negative" associations through momentum contrastive learning, the model effectively reduces spurious correlations. The final comprehensive representations and optimized graph structure are then used to predict potential MDAs. Experimental results demonstrate that ISFNMDA outperforms existing methods, and case studies further validate its predictive capability. © 2013 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Genome-wide association study (GWAS) provides insights into the genomic basis of reproduction-related traits in Chouardia litardierei (Asparagaceae)

引用

BMC Plant Biology 2025年第1期25卷 1-25页

作者： Šarančić, Sara Laura Pleić, Nikolina Mitić, Damjan Križanović, Krešimir Surina, Boštjan Radosavljević, Ivan Department of Biology Faculty of Science University of Zagreb Marulićev Trg 9 A Zagreb 10000 Croatia Department of Biology and Human Genetics School of Medicine University of Split Šoltanska 2 Split 21000 Croatia Department of Electronic Systems and Information Processing Faculty of Electrical Engineering and Computing University of Zagreb Unska 3 Zagreb 10000 Croatia Natural History Museum Rijeka Lorenzov Prolaz 1 Rijeka 51000 Croatia Faculty of Mathematics Natural Sciences and Information Technologies University of Primorska Glagoljaška 8 Koper 6000 Slovenia

Background: Chouardia litardierei, commonly known as amethyst meadow squill, is a plant species characterized by profound ecological plasti vcity. As a wild, non-model species, it represents a suitable system for gaining insights into the genomic background of the local adaptation process. By implementing a genome-environment and genome-wide association studies, we sought to investigate the genomic regions related to the local adaptation and the development of several reproduction-related traits in C. litardierei: for sexual reproduction, Average Height of Inflorescences (AHI) and Total Flower Count (TFC) per genotype, and for asexual reproduction, Bulb Count (BC) per genotype. Results: A genome-environment association (GEA) study of selected C. litardierei populations revealed the precipitation of the coldest quarter as the bioclimatic variable with the most substantial influence on detected variability, with numerous candidate genes detected and functionally characterized. To evaluate the genetic basis of selected reproduction-related traits we combined phenotypic data of 214 individuals raised as a part of a common garden experiment with ddRADseq genotyping results. After implementing various single- and multi-locus GWAS models for all traits, multiple candidate loci affecting their development were recognized. In addition, high, narrow-sense heritability estimates indicated that genetic factors accounted for over 55% of the phenotypic variance in each trait. Notably, the highest heritability estimate was observed for the Average Height of Inflorescences (71.95%), suggesting its crucial role in reproductive success. Functional annotation of the associated genomic regions identified key protein families involved in reproduction-related biological pathways, including nitrogen metabolism, phytohormone regulation, and floral organs development. Conclusion: By implementing GEA and GWAS, we revealed a list of candidate loci significantly associated with adaptation to s

关键词： Adaptive traits Chouardia litardierei GEA GWAS Local adaptation Reproduction

来源：评论

学校读者我要写书评

暂无评论

A Survey on Decentralized Identifiers and Verifiable Credentials

引用

IEEE Communications Surveys and Tutorials 2025年

作者： Mazzocca, Carlo Acar, Abbas Uluagac, Selcuk Montanari, Rebecca Bellavista, Paolo Conti, Mauro University of Salerno Department of Information and Electrical Engineering and Applied Mathematics Fisciano84084 Italy Florida International University Cyber-Physical Systems Security Lab School of Computing and Information Science MiamiFL33174 United States University of Bologna Department of Computer Science and Engineering Bologna40136 Italy University of Padua Department of Mathematics Padua35131 Italy

Digital identity has always been one of the keystones for implementing secure and trustworthy communications among parties. The ever-evolving digital landscape has undergone numerous technological transformations that have profoundly reshaped digital identity management, leading to a major shift from centralized to decentralized identity models. The latest stage of this evolution is represented by the emerging paradigm of Self-Sovereign Identity (SSI), which gives identity owners full control over their data. SSI leverages Decentralized Identifiers (DIDs) and Verifiable Credentials (VCs), which have been recently standardized by theWorldWideWeb Consortium (W3C). These technologies have the potential to build more secure and decentralized digital identity systems, significantly strengthening communication security in scenarios involving many distributed participants. It is worth noting that use of DIDs and VCs is not limited to individuals but extends to a wide range of entities including cloud, edge, and Internet of Things (IoT) resources. However, due to their novelty, existing literature lacks a comprehensive survey on DIDs and VCs beyond the scope of SSI. This paper fills this gap by providing a comprehensive overview of DIDs and VCs from multiple perspectives. It identifies key security threats and mitigation strategies, analyzes available implementations to guide practitioners in making informed decisions, and reviews the adoption of these technologies across various application domains. Moreover, it also examines related regulations, projects, and consortiums emerging worldwide. Finally, it discusses the primary challenges hindering their real-world adoption and outlines future research directions. © 1998-2012 IEEE.

关键词： Decentralized control

来源：评论

学校读者我要写书评

暂无评论

QoS-Based Beamforming and Compression Design for Cooperative Cellular Networks via Lagrangian Duality

引用

IEEE Transactions on Signal Processing 2025年

作者： Fan, Xilai Liu, Ya-Feng Liu, Liang Chang, Tsung-Hui Chinese Academy of Sciences State Key Laboratory of Scientific and Engineering Computing Institute of Computational Mathematics and Scientific/Engineering Computing Academy of Mathematics and Systems Science Beijing100190 China The Hong Kong Polytechnic University Department of Electronic and Information Engineering Hong Kong Hong Kong The Chinese University of HongKong School of Science and Engineering Shenzhen Shenzhen China Shenzhen Research Institute of Big Data China

This paper considers the quality-of-service (QoS)-based joint beamforming and compression design problem in the downlink cooperative cellular network, where multiple relay-like base stations (BSs), connected to the central processor via ratelimited fronthaul links, cooperatively transmit messages to the users. The problem of interest is formulated as the minimization of the total transmit power of the BSs, subject to all users' signal-to-interference-plus-noise ratio (SINR) constraints and all BSs' fronthaul rate constraints. In this paper, we first show that there is no duality gap between the considered joint optimization problem and its Lagrangian dual by showing the tightness of its semidefinite relaxation (SDR). Then, we propose an efficient algorithm based on the above duality result for solving the considered problem. The proposed algorithm judiciously exploits the special structure of an enhanced Karush-Kuhn-Tucker (KKT) conditions of the considered problem and approaches the solution that satisfies the enhanced KKT conditions via two fixed point iterations. Two key features of the proposed algorithm are: (1) it is able to detect whether the considered problem is feasible or not and find its globally optimal solution when it is feasible;(2) it is highly efficient because both of the fixed point iterations in the proposed algorithm are linearly convergent and function evaluations in the fixed point iterations are computationally cheap. Numerical results show the global optimality and efficiency of the proposed algorithm. © 1991-2012 IEEE.

关键词： Lagrange multipliers

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：