检索结果-内蒙古大学图书馆

12th International Conference on Learning Representations, ICLR 2024

作者： Zhang, Yingtao Bai, Haoli Lin, Haokun Zhao, Jialin Hou, Lu Cannistraci, Carlo Vittorio Center for Complex Network Intelligence Tsinghua Laboratory of Brain and Intelligence China Department of Computer Science Tsinghua University China Department of Biomedical Engineering Tsinghua University China Huawei Noah's Ark Lab Canada Institute of Automation Chinese Academy of Sciences China

With the rapid growth of large language models (LLMs), there is increasing demand for memory and computation in LLMs. Recent efforts on post-training pruning of LLMs aim to reduce the model size and computation requirements, yet the performance is still sub-optimal. In this paper, we present a plug-and-play solution for post-training pruning of LLMs. The proposed solution has two innovative components: 1) Relative Importance and Activations (RIA), a new pruning metric that jointly considers the weight and activations efficiently on LLMs;and 2) Channel Permutation, a new approach to maximally preserve important weights under N:M sparsity. The two proposed components can be readily combined to further enhance the N:M semi-structured pruning of LLMs. Our empirical experiments show that RIA alone can already surpass all existing post-training pruning methods on prevalent LLMs, e.g., LLaMA ranging from 7B to 65B. Furthermore, N:M semi-structured pruning with channel permutation can even outperform the original LLaMA2-70B on zero-shot tasks, together with practical speedup on specific hardware. Our code is available at: https://***/biomedicalcybernetics/Relative-importance-and-activation-pruning. © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.

关键词： Chemical activation

来源：评论

学校读者我要写书评

暂无评论

A Design Space of Behavior Change Interventions for Responsible Data science 25

A Design Space of Behavior Change Interventions for Responsi...

引用

30th International Conference on Intelligent User Interfaces, IUI 2025

作者： Dong, Ziwei Barrett, Teanna Patil, Ameya B Shoda, Yuichi Battle, Leilani Wall, Emily Computer Science Department Emory University AtlantaGA United States Computer Science and Engineering/Social Futures Lab University of Washington SeattleWA United States

ISBN: (纸本)9798400713064

Behavior change theories, rooted in psychology and sociology, offer valuable insights into why and how individuals and groups modify their actions and decisions. By leveraging these theories in the context of responsible data science, we can better understand and influence the behaviors of data scientists, who play a central role in ensuring ethical outcomes by collecting data, developing, and deploying models. In this paper, we present a comprehensive design space for behavior change interventions aimed at promoting responsible behaviors in data science, structured around the 5W1H interrogative framework (Why, Who, What, When, Where, and How). This framework provides a practical guide for developing effective interventions designed to promote responsible behaviors in data science. We showcase the usability of this design space by using it to characterize existing responsible data science intervention tools. We further demonstrate its utility through two usage scenarios to show how the design space can be applied during the ideation phase for building effective tools to foster responsible data science practices. Our work equips the data science community with resources to create effective interventions that not only ensure technical excellence but also foster ethical responsibility, ultimately benefiting society through the responsible use of data. © 2025 Copyright held by the owner/author(s).

关键词： Data science

来源：评论

学校读者我要写书评

暂无评论

BD-Trust: behavioural and data trust management scheme for internet of things

引用

Journal of Ambient intelligence and Humanized Computing 2023年第12期14卷 16195-16207页

作者： Sharma, Avani Pilli, Emmanuel S. Mazumdar, Arka P. Department of Computer Science and Engineering Malaviya National Institute of Technology Jaipur India

The emergence of Internet of Things (IoT) has paved the way to realize ubiquitous access to manifold services and applications. Effective utilization of these services vitally demands secure deployment of IoT systems where data or information can be transmitted without worrying about the uncertainty of different devices in the system. Trust Management Systems (TMS) have proved themself as impressive solutions to attain this security requirement where entities are judged based on their behavior. For IoT, influential deployment of TMS requires coping with characteristics of IoT systems such as device heterogeneity, context-awareness, dynamicity, etc. Paying attention to the quality of received data for reliable service composition and accurate assessment is also a matter of concern. This paper presents a direct observation-based distributed trust management scheme, called BD-Trust, in which trust is measured by integrating behavioral trust of devices with the data trust. For evaluation of behavioral trust, BD-Trust reckons on three trust parameters, namely success rate, satisfaction level, and credibility. Data trust is measured considering consistency in received data values for the successful interaction responses. BD-Trust employs a trust update mechanism to alleviate the effect of malicious devices that try to disrupt the process of trust management. An adaptive decay factor has been designed to reflect previously observed trust value in recent computation while updating trust scores. The effectiveness of BD-Trust has been validated through simulation experiments and results have been compared with two state-of-art schemes for trust convergence, assessment accuracy, and resilience against attacks, considering a dynamic environment. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

ChatCite: LLM Agent with Human Workflow Guidance for Comparative Literature Summary 31

ChatCite: LLM Agent with Human Workflow Guidance for Compara...

引用

31st International Conference on Computational Linguistics, COLING 2025

作者： Li, Yutong Chen, Lu Liu, Aiwei Yu, Kai Wen, Lijie Tsinghua University Beijing China X-LANCE Lab Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence SJTU AI Institute Shanghai Jiao Tong University Shanghai China Suzhou Laboratory Suzhou China

ISBN: (纸本)9798891761964

The literature review is an indispensable step in the research process. It provides the benefit of comprehending the research problem and understanding the current research situation while conducting a comparative analysis of prior works. However, literature summary is challenging and time consuming. The previous LLM-based studies on literature review mainly focused on the complete process, including literature retrieval, screening, and summarization. However, for the summarization step, simple CoT method often lacks the ability to provide extensive comparative summary. In this work, we firstly focus on the independent literature summarization step and introduce ChatCite, an LLM agent with human workflow guidance for comparative literature summary. This agent, by mimicking the human workflow, first extracts key elements from relevant literature and then generates summaries using a Reflective Incremental Mechanism. In order to better evaluate the quality of the generated summaries, we devised a LLM-based automatic evaluation metric, G-Score, in refer to the human evaluation criteria. The ChatCite agent outperformed other models in various dimensions in the experiments. The literature summaries generated by ChatCite can also be directly used for drafting literature *** code is available. © 2025 Association for Computational Linguistics.

关键词： Chatbots

来源：评论

学校读者我要写书评

暂无评论

Blockchain-powered Healthcare: Revolutionizing Security and Privacy in IoT-based Systems

Blockchain-powered Healthcare: Revolutionizing Security and ...

引用

2024 International Conference on Computational intelligence and Computing Applications, ICCICA 2024

作者： Singla, Deepak Kumar, Sanjeev Dhingra, Deepti Ghandhi, Anju Department Of Computer Science And Engineering Ambala Haryana Mullana India Panipat Institute Of Engineering And Technology Department Of Computer Science And Engineering Panipat Haryana Samalkha India

ISBN: (纸本)9798350306446

The rapid growth of the IoT and wearable device industries has expanded remote patient monitoring, but healthcare systems face security vulnerabilities inherent in their client/server architecture. To address these challenges and meet regulatory standards, a shift to a distributed architecture is necessary. Blockchain technology offers promise with its distributed nature and security features. MediSecureNet, proposed in this paper, integrates blockchain and IoT to create a secure healthcare system. Prioritizing security, scalability, and processing speed, MediSecureNet focuses on remote patient monitoring, particularly for chronic conditions. It incorporates a re-encryption proxy for security, blockchain for hash data storage, and smart contracts for access management. Scalability is achieved through an off-chain IPFS database, and an Ethereum-based proof of authority mechanism enhances data storage efficiency. © 2024 IEEE.

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

Design and Development of a Device to Measure Tissue Strength in Experimental Animals 8

Design and Development of a Device to Measure Tissue Strengt...

引用

8th IEEE International Conference on distributed Computing, VLSI, electrical Circuits and Robotics, DISCOVER 2024

作者： Veeresha, R.K. Shastry, C.S. Moolya Shodhan, S. Karegoudra, Shilpa Kumar, Nithin Department of Robotics and Artificial Intelligence Udupi574110 India Paneer Deralakatte Mangalore575018 India Department of Mechanical Engineering Udupi574110 India Department of Computer Science and Engineering Udupi574110 India

ISBN: (纸本)9798350350593

Developing a wound strength measuring device for experimental animals involves a multidisciplinary approach that combines engineering, biology, and veterinary medicine. Such a device would help researchers assess the healing progress and strength of wounds in animals, enabling a better understanding of tissue regeneration and wound healing mechanisms. In this paper, We designed a mice holder along with a full experimental setup to apply load and measure the displacement automatically. Also, we carried out an experiment to measure load versus displacement. © 2024 IEEE.

关键词： Mammals

来源：评论

学校读者我要写书评

暂无评论

Federated Clustering Defense: Mitigating Data Exposure in Collaborative Learning

Federated Clustering Defense: Mitigating Data Exposure in Co...

引用

2025 International Conference on Emerging Technologies for Intelligent Systems, ETIS 2025

作者： Radhakrishnan, Anuja Krishna, Sathya Bama Sathyabhama Institute of Science & Technology Department of Computer Science and Engineering Chennai India

ISBN: (纸本)9798331507541

The development of intelligent industrial applications is being driven globally by advancements in the next generation of Internet of Things (IoT) technologies. At the same time, artificial intelligence (AI) technologies like machine learning and deep learning are becoming more widely used. Conventional machine learning models primarily rely on large data sets. However, the vast data sets generated by network-edge devices are expensive, inefficient to acquire and handle, and they pose major privacy hazards. Federated learning (FL) is a novel paradigm for statistical model training on distributed edge networks that allows data to take part in the training of federated models without being localized. Traditional machine learning issues with low data usage, data privacy, and information security brought on by data isolation can be resolved with this method. We propose a novel Federated Clustering with Differential Privacy (FedCDP) algorithm designed to efficiently model user behavior in an unsupervised manner. This algorithm focuses on optimizing computation and storage requirements while minimizing data leakage through robust differential privacy mechanisms. © 2025 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Al-Biruni Earth Radius(BER)Metaheuristic Search Optimization Algorithm

引用

computer Systems science & engineering 2023年第5期45卷 1917-1934页

作者： El-Sayed M.El-kenawy Abdelaziz A.Abdelhamid Abdelhameed Ibrahim Seyedali Mirjalili Nima Khodadad Mona A.Al duailij Amel Ali Alhussan Doaa Sami Khafaga Department of Communications and Electronics Delta Higher Institute of Engineering and TechnologyMansoura35111Egypt Faculty of Artificial Intelligence Delta University for Science and TechnologyMansoura35712Egypt Department of Computer Science Faculty of Computer and Information SciencesAin Shams UniversityCairo11566Egypt Department of Computer Science College of Computing and Information TechnologyShaqra University11961Saudi Arabia Computer Engineering and Control Systems Department Faculty of EngineeringMansoura UniversityMansoura35516Egypt Centre for Artificial Intelligence Research and Optimization Torrens University AustraliaFortitude ValleyQLD 4006Australia Yonsei Frontier Lab Yonsei UniversitySeoul03722Korea Department of Civil and Environmental Engineering Florida International UniversityMiamiUSA Department of Computer Sciences College of Computer and Information SciencesPrincess Nourah Bint Abdulrahman UniversityP.O.Box 84428Riyadh11671Saudi Arabia

Metaheuristic optimization algorithms present an effective method for solving several optimization problems from various types of applications and *** metaheuristics and evolutionary optimization algorithms have been emerged recently in the literature and gained widespread attention,such as particle swarm optimization(PSO),whale optimization algorithm(WOA),grey wolf optimization algorithm(GWO),genetic algorithm(GA),and gravitational search algorithm(GSA).According to the literature,no one metaheuristic optimization algorithm can handle all present optimization *** novel optimization methodologies are still *** Al-Biruni earth radius(BER)search optimization algorithm is proposed in this *** proposed algorithm was motivated by the behavior of swarm members in achieving their global *** search space around local solutions to be explored is determined by Al-Biruni earth radius calculation method.A comparative analysis with existing state-of-the-art optimization algorithms corroborated the findings of BER’s validation and testing against seven mathematical optimization *** results show that BER can both explore and avoid local *** has also been tested on an engineering design optimization *** results reveal that,in terms of performance and capability,BER outperforms the performance of state-of-the-art metaheuristic optimization algorithms.

关键词： Metaheuristics evolutionary optimization exploration exploitation mutation Al-biruni earth radius

来源：评论

学校读者我要写书评

暂无评论

Next-Generation IoT Intrusion Detection: Leveraging Hybrid Deep Learning Algorithms for Enhanced Security 2

Next-Generation IoT Intrusion Detection: Leveraging Hybrid D...

引用

2nd International Conference on Computational intelligence, Communication Technology and Networking, CICTN 2025

作者： Urvashi Agarwal, Parul Raghuvanshi, Kamlesh Kumar Ahmed, Jawed Department of Computer Science and Engineering Jamia Hamdard University Delhi India Department of Computer Science University of Delhi Delhi India

ISBN: (纸本)9798331530389

Wireless sensor node and DDoS (distributed Denial of Service) is a major attack way that can exhaust the energy of a sensor, disrupt the communication traffic of an IoT network. We must be on the lookout for these attacks. With the quantity of sensor nodes, traditional methods face difficulty in detecting network traffic analysis requiring big data analysis. This paper is dedicated to Network Traffic analysis using Big Data environment and detecting attacks using Deep Learning. Implemented with PySpark in Google Colab, using Keras and Scikit-Learn built-in libraries. To be more specific, training and testing are conducted on the 'CICIoT2023'/'TON IoT' datasets, respectively. When we use a correlation feature reducer, it virtually considers the most significant features for inclusion and reduces all of our features. The first contains a hybrid deep leaning model combining one-dimensional EfficientDet and Gradient Boosting Machine (GBM), comparing it with ten other algorithms. It gen- erated 99.95% binary classification accuracy for the CICIoT2023 dataset and 99.75% for TON IoT data. © 2025 IEEE.

关键词： Network intrusion

来源：评论

学校读者我要写书评

暂无评论

Cloud Based Smart Medical Recommendation System Using Pattern Recognition 1

Cloud Based Smart Medical Recommendation System Using Patter...

引用

1st IEEE International Conference on Networking and Communications, ICNWC 2023

作者： Sujatha, E. Shanthasheela, S. Gopirajan, P.V. Department of Computer Science and Engineering Thandalam Chennai602105 India Velammal Institute of Technology Department of Computer Science and Engineering Chennai India SRM Institute of Science and Technology School of Computing Department of Computational Intelligence Kattankulathur Chennai603203 India

ISBN: (纸本)9798350336009

To build smart medical recommendation system the data needs to be collected from laboratories. The system is intended to provide services as control and manage data samples, industry standards, real test results, reports, human resources in the laboratories, machines, tiny instruments, and automation of the workflow. Combining the data from the laboratories and medical institutions enable ease of communication without time delay. The challenging task is that identifying appropriate hardware and software components which provides compatibility in collecting data and their processing steps, repository storage, secured distribution, and the policies used for information representation in the results. In order to automate the entire process of medical recommendation system the metric parameters are used for determining the efficiency and productivity of the lab to enhance the quality of control systems. The vulnerability lies in maintaining traditional way of records in paper and error in the data. The purpose of this project is to develop the process of daily tasks will be computed through the database using pattern recognition techniques. © 2023 IEEE.

关键词： Recommender systems

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

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

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：