Computer Standards & Interfaces最新文献

筛选
英文 中文
Survey of reversible data hiding: Statistics, current trends, and future outlook
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-25 DOI: 10.1016/j.csi.2025.104003
Sonal Gandhi, Rajeev Kumar
{"title":"Survey of reversible data hiding: Statistics, current trends, and future outlook","authors":"Sonal Gandhi,&nbsp;Rajeev Kumar","doi":"10.1016/j.csi.2025.104003","DOIUrl":"10.1016/j.csi.2025.104003","url":null,"abstract":"<div><div>In the era of increasing digital media storage and transmission over networks, reversible data hiding (RDH) has evolved as a prominent area of research mitigating information security risk. To study the evolution of research, highlight its achievements over the years, and provide future prospects, this paper presents an extensive review of RDH utilizing the dataset extracted from one of the most popular and exhaustive databases, Web of Science. The study aims to first perform quantitative analysis that includes trend analysis, citation analysis, prominent authors and organizations, and geographical coverage, along with qualitative analysis focusing on key research areas and future prospects within RDH. The study further provides a structured view of sub-technologies within RDH, along with the key contributors and their proposed techniques that have led to the evolution of RDH over the years. Next, we provide a comprehensive review of some of the prominent works in each of the sub-technologies of RDH. Finally, several key research directions, identified based on current research trends and early-stage problems and motivations, are discussed. Overall, the proposed study provides valuable insights into the evolution, key milestones, current state, and future prospects of RDH, serving as a guide for the research community.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 104003"},"PeriodicalIF":4.1,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143725098","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
PP-MAD: Privacy-preserving multi-task data aggregation in mobile crowdsensing via blockchain
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-25 DOI: 10.1016/j.csi.2025.104002
Xingfu Yan , Jiaju Ding , Fucai Luo , Zheng Gong , Wing W.Y. Ng , Yiyuan Luo
{"title":"PP-MAD: Privacy-preserving multi-task data aggregation in mobile crowdsensing via blockchain","authors":"Xingfu Yan ,&nbsp;Jiaju Ding ,&nbsp;Fucai Luo ,&nbsp;Zheng Gong ,&nbsp;Wing W.Y. Ng ,&nbsp;Yiyuan Luo","doi":"10.1016/j.csi.2025.104002","DOIUrl":"10.1016/j.csi.2025.104002","url":null,"abstract":"<div><div>In smart city, multi-task data aggregation has become a key method for extracting useful information from massive sensing data generated by concurrent mobile crowdsensing tasks from multiple task requesters. In such multi-requester and multi-task scenario, each task requester wants to protect the privacy of their own aggregation results. Thus, protecting privacies of both workers and task requesters pose a significant challenge for multi-task data aggregation. Most existing privacy-preserving data aggregation methods focus on single-requester scenarios. When applied to multi-task and multi-requester aggregation, existing methods are inefficient due to completing repeatedly each task and fail to safeguard the privacy of each task requester. Additionally, existing multi-task data aggregation schemes do not support multiple types of aggregation. To tackle these challenges, we propose PP-MAD, a multi-type and privacy-preserving multi-task data aggregation scheme based on blockchain for mobile crowdsensing. PP-MAD is able to aggregate multiple concurrent tasks from multiple task requesters, and it supports many types of data aggregation, including sum, mean, variance, weighted sum, weighted mean. Moreover, PP-MAD ensures privacies of workers’ data and aggregation results of each task requester, even under collusion attacks. A detailed security analysis verifies that PP-MAD is both secure and privacy-preserving. Furthermore, experimental results and theoretical analyses of both computation and communication costs demonstrate our scheme is efficient.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 104002"},"PeriodicalIF":4.1,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143704075","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
PQHealthcare: An efficient cloud data sharing system via Functional Encryption from lattices
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-19 DOI: 10.1016/j.csi.2025.103993
Qiqi Lai , Chongshen Chen , Momeng Liu , Yang Yang , Yong Yu
{"title":"PQHealthcare: An efficient cloud data sharing system via Functional Encryption from lattices","authors":"Qiqi Lai ,&nbsp;Chongshen Chen ,&nbsp;Momeng Liu ,&nbsp;Yang Yang ,&nbsp;Yong Yu","doi":"10.1016/j.csi.2025.103993","DOIUrl":"10.1016/j.csi.2025.103993","url":null,"abstract":"<div><div>With the rapid development of the large language models (<span><math><mi>LLM</mi></math></span>) service on the cloud, especially in the healthcare sector, the usage of large models is becoming increasingly popular. However, training these models always involves sensitive information containing lots of personal medical data. And thus, such training processes might result in the exposure of privacy information. In order to help the users eliminate their concerns and share their data in a secure way, we need to find a privacy-preserving method for data sharing in the public cloud service environment.</div><div>While traditional public-key encryption (<span><math><mi>PKE</mi></math></span>) schemes can effectively encrypt healthcare data, they typically offer protection in an “all-or-nothing” manner, lacking flexibility and imposing a significant computational burden on public cloud servers. In contrast, Functional Encryption (<span><math><mi>FE</mi></math></span>) offers a more flexible way of encryption with access control, making it ideal for the cloud data sharing environment. Furthermore, we observe that the majority of computations involved in training large models can be represented by inner product functions. To establish a secure public cloud data sharing system, we propose an efficient <span><math><mi>FE</mi></math></span> scheme for inner product function class, making it well-suited for various real-world applications. We will demonstrate that our scheme achieves post-quantum security based on lattice assumptions.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 103993"},"PeriodicalIF":4.1,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143684751","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Standardization in artificial general intelligence model for education
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-19 DOI: 10.1016/j.csi.2025.104006
Qiuxuan Xu , Yonghe Wu , Hao Zheng , Huan Yan , Huina Wu , Yu Qian , You Wu , Bowen Liu
{"title":"Standardization in artificial general intelligence model for education","authors":"Qiuxuan Xu ,&nbsp;Yonghe Wu ,&nbsp;Hao Zheng ,&nbsp;Huan Yan ,&nbsp;Huina Wu ,&nbsp;Yu Qian ,&nbsp;You Wu ,&nbsp;Bowen Liu","doi":"10.1016/j.csi.2025.104006","DOIUrl":"10.1016/j.csi.2025.104006","url":null,"abstract":"<div><div>The application of Artificial General Intelligence Models (AGIMs) in education has been identified as a promising emerging field. However, extensive research has revealed limitations in using AGIM in education, particularly in terms of controllability, trustworthiness, explainability, evaluation and feedback, security, and privacy. Therefore, standardization in AGIMs for Education (AGIME) is urgently required to provide normative guidance for developing artificial intelligence systems in education. This study first explores an AGIME standardization process with the methodology of use case collection and iterative research. We then propose the definition and attributes of AGIME and establish a standard system framework for the AGIME life cycle. This framework includes published specifications such as information model, data specification, evaluation specification, and application requirements on teaching and learning. We introduce standard application cases to validate the effectiveness of AGIME standard system framework. Finally, we present several specifications currently under development within this standard system, including interface, regulatory, operation and maintenance, and security, ethics, and privacy specifications. This study provides references for AGIME development and deployment, ensuring the technical stability, data credibility, evaluation accuracy, and pedagogical applicability of AGIME.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 104006"},"PeriodicalIF":4.1,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143725179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
A privacy-preserving standardized model for large-scale source code fingerprint extraction and clone detection
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-19 DOI: 10.1016/j.csi.2025.103998
Ming Yang , Yu-an Tan , Ning Shi , Yajie Wang , Ziqi Wang , Qi Liang
{"title":"A privacy-preserving standardized model for large-scale source code fingerprint extraction and clone detection","authors":"Ming Yang ,&nbsp;Yu-an Tan ,&nbsp;Ning Shi ,&nbsp;Yajie Wang ,&nbsp;Ziqi Wang ,&nbsp;Qi Liang","doi":"10.1016/j.csi.2025.103998","DOIUrl":"10.1016/j.csi.2025.103998","url":null,"abstract":"<div><div>With the rapid advancement of software technology, developers often replicate or modify existing code to achieve code cloning, thereby improving development efficiency. However, the widespread use of open-source code may lead to intellectual property disputes and infringement risks. Additionally, the repeated use of cloned code can exacerbate vulnerabilities, increasing system fragility and maintenance costs, especially when synchronized modifications are required for cloned fragments during software evolution. To address these challenges, this paper proposes a privacy-preserving large-scale code fingerprint extraction model—Ringer. The model decouples feature extraction from clone detection, enabling efficient clone detection without direct access to the source code. Ringer employs syntax trees for lexical and syntactic analysis to comprehensively extract code features, and generates irreversible code fingerprints through token replacement and the Metro-128 hash algorithm, ensuring the privacy of the source code while effectively detecting clones. Experimental results show that Ringer performs excellently on datasets from multiple programming languages (e.g., Java, C++, Python, etc.), maintaining consistently high accuracy based on the characteristics of each language. On the Python dataset, Ringer achieves detection accuracies of 94%, 94%, and 97% for Type-1, Type-2, and Type-3 clones, respectively, further validating its efficiency and reliability in practical applications. Compared to mainstream detection tools (e.g., Moss and NiCad), Ringer outperforms in cross-language detection, demonstrating its robust adaptability and superior accuracy. This strongly supports the broad applicability of Ringer for privacy-preserving clone detection in large-scale, multi-language codebases.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 103998"},"PeriodicalIF":4.1,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143704124","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Pypen: Code instrumentation tool for dynamic analysis and energy efficiency evaluation
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-19 DOI: 10.1016/j.csi.2025.104000
Carlos Pulido, Ignacio García, Ma Ángeles Moraga, Félix García, Coral Calero
{"title":"Pypen: Code instrumentation tool for dynamic analysis and energy efficiency evaluation","authors":"Carlos Pulido,&nbsp;Ignacio García,&nbsp;Ma Ángeles Moraga,&nbsp;Félix García,&nbsp;Coral Calero","doi":"10.1016/j.csi.2025.104000","DOIUrl":"10.1016/j.csi.2025.104000","url":null,"abstract":"<div><div>Given the large environmental impact that software can generate, the environmental sustainability of software is increasingly a major concern for developers today. Thus, it is important to provide them with the means and tools to understand the impact of their decisions on the energy consumption of software. To achieve this, more accurate information must be made available to better guide optimization efforts, in particular by identifying the specific sections of the software where the highest energy consumption occurs.</div><div>This paper presents Pypen, a profiler that uses instrumentation to collect detailed execution data about the software execution. Through its integration with the ecosystem of tools of the technological component of the FEETINGS framework, Pypen is able to identify “hot” methods where the highest energy consumption occurs, allowing developers to focus on improving the energy efficiency of specific software sections.</div><div>To evaluate and demonstrate the effectiveness of our tool, a sample study was conducted aimed at identifying areas of the software with higher energy consumption, providing opportunities to better focus improvement efforts.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 104000"},"PeriodicalIF":4.1,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143684749","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Revealing the mobile UX horizon: Exploring user experience aspects, attributes, and measurement methods - A systematic mapping study
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-16 DOI: 10.1016/j.csi.2025.103999
Haifa Alshammare , Mohammad Alshayeb , Malak Baslyman
{"title":"Revealing the mobile UX horizon: Exploring user experience aspects, attributes, and measurement methods - A systematic mapping study","authors":"Haifa Alshammare ,&nbsp;Mohammad Alshayeb ,&nbsp;Malak Baslyman","doi":"10.1016/j.csi.2025.103999","DOIUrl":"10.1016/j.csi.2025.103999","url":null,"abstract":"<div><h3>Context</h3><div>The exponential growth of mobile internet usage underscores the critical importance of investigating mobile app enhancements across various domains. User Experience (UX) emerges as a pivotal quality factor, encompassing emotional, sensory, and contextual dimensions, significantly influencing user behavior and retention.</div></div><div><h3>Objective</h3><div>This systematic mapping study investigates research efforts on mobile application UX to discern domains where UX has been explored, identifying UX aspects, attributes, and assessment methods.</div></div><div><h3>Method</h3><div>Through a standard systematic mapping procedure, 189 primary studies were identified from 9107 reviewed articles across five digital libraries.</div></div><div><h3>Results</h3><div>100 distinct UX aspects and 166 attributes were identified and categorized into App, User, and Context aspects. These aspects were distributed among 14 domains, with healthcare being extensively researched. The usability aspect is prominent in 54.5 % of studies. Efficiency, satisfaction, effectiveness, learnability, memorability, attractiveness, errors, and ease of use emerged as the most commonly utilized attributes—notably, these attributes are considered usability attributes among several standards. Additionally, 38 and 22 evaluation methods were identified for aspects and attributes, grouped into self-reported, physiological, observational, and expert evaluation categories. Questionnaires were predominantly utilized.</div></div><div><h3>Conclusion</h3><div>This review summarizes the mobile application UX landscape, highlighting key aspects, attributes, and evaluation methods across domains. Future research should leverage these insights to advance UX design and evaluation practices, fostering innovation in software engineering and human-computer interaction.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 103999"},"PeriodicalIF":4.1,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143684750","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
A blockchain and signature based scheme for cross-domain authentication with decentralized identity
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-13 DOI: 10.1016/j.csi.2025.103994
Zinuo Zhang , Wei Ren , Xianchao Zhang , Yani Sun , Tianqing Zhu , Kim-Kwang Raymond Choo
{"title":"A blockchain and signature based scheme for cross-domain authentication with decentralized identity","authors":"Zinuo Zhang ,&nbsp;Wei Ren ,&nbsp;Xianchao Zhang ,&nbsp;Yani Sun ,&nbsp;Tianqing Zhu ,&nbsp;Kim-Kwang Raymond Choo","doi":"10.1016/j.csi.2025.103994","DOIUrl":"10.1016/j.csi.2025.103994","url":null,"abstract":"<div><div>Currently, users access various network services without a unified identity authentication among domains, which not only obliges users to numerous network accounts, but also creates significant inconvenience for cross-domain authentication. Most existing schemes rely on the Public Key Infrastructure (PKI) system, which depends on the trustworthiness of Certificate Authorities (CAs). However, this poses challenges to maintaining user identity privacy. Addressing this issue, this paper proposes a blockchain-based cross-domain identity authentication scheme utilizing decentralized identity management, where users only need to register an account within one domain and can access services in other domains through verifiable credentials (VC). The scheme designs two types of VC, namely, directional VCs and general VCs. General VCs use threshold signatures allowing access to a larger number of domains. It also introduces proxy signatures to design a method for credential borrowing, further enhancing user convenience. Logical analysis using the Subject–Verb–Object (SVO) structure confirms the scheme’s viability, while security analysis indicates its resilience against various attacks, including replay, impersonation, and internal threats. Moreover, the scheme provides substantial benefits for safeguarding user privacy and data security.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 103994"},"PeriodicalIF":4.1,"publicationDate":"2025-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143628548","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
A dynamic Bayesian network risk assessment model for coal-fired power plants based on grey correlation and triangular fuzzy theory
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-12 DOI: 10.1016/j.csi.2025.104001
Yinqing Wang , Xiangchun Li , Chunli Yang , Tao Yang , Fengchun Lan , Xin Tian
{"title":"A dynamic Bayesian network risk assessment model for coal-fired power plants based on grey correlation and triangular fuzzy theory","authors":"Yinqing Wang ,&nbsp;Xiangchun Li ,&nbsp;Chunli Yang ,&nbsp;Tao Yang ,&nbsp;Fengchun Lan ,&nbsp;Xin Tian","doi":"10.1016/j.csi.2025.104001","DOIUrl":"10.1016/j.csi.2025.104001","url":null,"abstract":"<div><div>This article adopts analysis methods such as Bayesian network model, grey correlation analysis, analytic hierarchy process, and three-level fuzzy method to establish a risk assessment index system and dynamic risk assessment model for coal-fired power plants. The results show that starting from the perspective of \"human machine environment management\",relatively complete risk assessment index system for coal-fired power plants has been determined using the grey correlation analysis method, with a total of 4 first level indicators, 15 second level indicators, and 63 third level indicators; Using Analytic Hierarchy Process to determine the weights of indicators at all levels, the calculation results are incorporated into subsequent Bayesian network models, transforming the Bayesian model from a static model to a dynamic model; Then bring the calculated results into NETICA software for calculation, obtain the original Bayesian network model, and perform reverse inference analysis and sensitivity analysis; Through comparative analysis, B1 (management personnel), B8 (equipment and facility management), B10 (work environment hazards), B14 (dual prevention mechanism management), X4 (personnel \"three violations\" situation), X20 (equipment change management), and X43 (vibration hazards) have significant changes. Therefore, management should be strengthened in practical operations, providing a theoretical basis for the actual production management of power plants.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 104001"},"PeriodicalIF":4.1,"publicationDate":"2025-03-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143684748","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Accelerating Mixture-of-Experts language model inference via plug-and-play lookahead gate on a single GPU
IF 4.1 2区 计算机科学
Computer Standards & Interfaces Pub Date : 2025-03-08 DOI: 10.1016/j.csi.2025.103996
Jie Ou, Yueming Chen, Buyao Xiong, Zhaokun Wang, Wenhong Tian
{"title":"Accelerating Mixture-of-Experts language model inference via plug-and-play lookahead gate on a single GPU","authors":"Jie Ou,&nbsp;Yueming Chen,&nbsp;Buyao Xiong,&nbsp;Zhaokun Wang,&nbsp;Wenhong Tian","doi":"10.1016/j.csi.2025.103996","DOIUrl":"10.1016/j.csi.2025.103996","url":null,"abstract":"<div><div>The widespread adoption of large language models (LLMs) has encouraged researchers to explore strategies for running these models more efficiently, such as the mixture of experts (MoE) method, which aims to increase the knowledge capacity of the model without substantially increasing its computational costs, as only a fraction of the model components are active for each token. However, this approach also increases the size of the model, which makes it challenging to run these models even on high-end GPUs. Quantization and offloading strategies have been used to enable the execution of MoE in resource-constrained environments, however, the time overhead introduced by offloading remains a bottleneck. In this paper, we propose a plug-and-play lookahead gate that predicts in advance the experts to be used in the next few layers. Furthermore, to mitigate the misalignment problem arising from cross-layer prediction, we introduce an alignment training method, layer-wise gate alignment, enhancing the prediction hit rate while maintaining low resource requirements. Moreover, we present a speculative expert scheduling strategy to accelerate the end-to-end inference process of MoE models. To validate our approach, we established an inference framework for quantized MoE and conducted extensive experiments. The results demonstrate the effectiveness of our proposed methods, with throughput improvements of 57.72%, 60.00%, and 62.26% under 4, 3, and 2-bit quantization conditions for experts, respectively, compared with the Mixtral-offloading method.</div></div>","PeriodicalId":50635,"journal":{"name":"Computer Standards & Interfaces","volume":"94 ","pages":"Article 103996"},"PeriodicalIF":4.1,"publicationDate":"2025-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143611213","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
相关产品
×
本文献相关产品
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信