检索结果-内蒙古大学图书馆

Clustering ensemble algorithm with high-order consistency learning

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Gan, Jianwen Chen, Yan Zhou, Peng Du, Liang School of Computer and Information Technology Shanxi University Shanxi Taiyuan030006 China College of Computer Science Sichuan University Sichuan Chengdu610065 China School of Computer Science and Technology Anhui University Anhui Hefei230601 China Institute of Big Data Science and Industry Shanxi University Shanxi Taiyuan030006 China

Most of the research on clustering ensemble focuses on designing practical consistency learning algorithms. To solve the problems that the quality of base clusters varies and the low-quality base clusters have an impact on the performance of the clustering ensemble, from the perspective of data mining, the intrinsic connections of data were mined based on the base clusters, and a high-order information fusion algorithm was proposed to represent the connections between data from different dimensions, namely Clustering Ensemble with High-order Consensus learning(HCLCE). Firstly, each high-order information was fused into a new structured consistency matrix. Then, the obtained multiple consistency matrices were fused together. Finally, multiple information was fused into a consistent result. Experimental results show that LCLCE algorithm has the clustering accuracy improved by an average of 7. 22%, and the Normalized Mutual Information(NMI) improved by an average of 9. 19% compared with the suboptimal Locally Weighted Evidence Accumulation (LWEA) algorithm. It can be seen that the proposed algorithm can obtain better clustering results compared with clustering ensemble algorithms and using one information alone. © 2024, CC BY.

关键词： Consensus algorithm

Federated Learning in STAR-RIS Aided SAGINs

学校读者我要写书评

暂无评论

IEEE Communications Standards Magazine

IEEE Communications Standards Magazine 2025年第2期9卷 23-29页

作者： Wu, Min Guo, Kefeng Dev, Kapal Hamadi, Hussam Al Singh, Keshav Khowaja, Sunder Ali Li, Xingwang Space Engineering University School of Space Information Beijing101407 China Munster Technological University CONNECT Centre Department of Computer Science Ireland College of Engineering and IT University of Dubai United Arab Emirates National Sun Yat-sen University Institute of Communications Engineering Kaohsiung Taiwan Technological University Dublin School of Computing Faculty of Computing Digital and Data Ireland Henan Polytechnic University School of Physics and Electronic Information Engineering Jiaozuo454000 China Jiaozuo Key Laboratory of Crow-Sensing Network Jiaozuo454003 China

Satellite-aerial-ground integrated network (SAGIN) is considered to be an advanced architecture that offers seamless coverage, flexible access, and efficient connectivity. However, several challenges emerge during the construction of SAGINs, which include complex propagation environments and limited spectrum resources. The aforementioned challenges can be addressed in a cost effective manner by integrating simultaneous transmission and reflecting reconfigurable intelligent surfaces (STAR-RISs). The STAR-RISs are capable of reflecting and transmitting the incident signals passively in order to serve users on both sides of the panel. Given the self-organizing nature of STAR-RIS-aided SAGINs, meticulous resource allocation and transmission strategy adaptation are imperative to accommodate changes in multiple communication node environments. Recently, federated learning has been envisioned as a pioneering distributed machine learning (ML) approach for managing the multi-objective optimization problem in STAR-RIS-aided SAGINs. This approach strives to balance the trade-offs between wireless communication performance and model efficiency. In this work, we present the advantages and potential directions for employing federated learning in STAR-RIS-aided SAGINs through multi-objectives optimization. Furthermore, to illustrate the practical application of federated learning in STAR-RIS-aided SAGINs, we present two case studies utilizing federated deep reinforcement learning, based on the non-orthogonal multiple access (NOMA) scheme. © 2017 IEEE.

关键词： Deep reinforcement learning

A Novel Stress Identification based on the Health Aspects of an Individual using Strategical Learning Methodology 2

学校读者我要写书评

暂无评论

A Novel Stress Identification based on the Health Aspects of...

2nd International Conference on Augmented Intelligence and Sustainable Systems, ICAISS 2023

作者： Periyasamy, P.N. Rajesh Kumar, K. Senthilvel, P. Gururama Duraimutharasan, N. Hannah Rose Esther, T. Hindusthan Institute of Technology Department of Computer Science and Engineering Coimbatore641032 India Chendu College of Engineering Department of Electronics and Communication Engineering Tamil Nadu Madurantakam603311 India Department of Computer Science and Engineering Saveetha School of Engineering Chennai602 105 India Amet University Department of Information Technology Tamil Nadu Kanathur603112 India Velammal Engineering College Department of Artificial Intelligence and Data Science Chennai600066 India

ISBN: (纸本)9798350325799

The modern human experience increasingly includes high levels of stress. According to the World Health Organization, stress is a mental illness that can negatively affect physical health. The effects of stress and depression can affect anyone, anywhere. There is tension in everyone's life. Stress is a major indicator of emotional well-being. Every area of a person's life, including their emotions, thoughts, and behaviors, are impacted by stress. Analyzing physiological signs is a traditional way to detect stress, according to earlier study. These readings come from sensors that are worn by the person being monitored. Researchers have attempted to detect stress by studying physiological information with traditional machine learning methods. The accuracy of the outcomes varies widely, from around 50% to almost 90%. Traditional machine learning methods have the issue of requiring handcrafted features. Accuracy drops if attributes are recognized wrongly. This research study has created a hybrid deep learning model to detect staff stress. With the intention of enhancing life quality, this study showed how deep neural networks may be used to create reliable, continuous, and noninvasive techniques for stress detection and emotion categorization. As they enable ready libraries for stress prediction, these ready tools and languages make it straightforward to predict stress. data science approaches are efficient in processing the training datasets and can accurately and quickly predict the human stress. © 2023 IEEE.

关键词： Stresses

Siamese Network Visual Tracking Algorithm Based on GCT Attention and Dual-Template Update 5

学校读者我要写书评

暂无评论

Siamese Network Visual Tracking Algorithm Based on GCT Atten...

5th International Conference on Natural Language Processing, ICNLP 2023

作者： Ma, Sugang Sun, Siwei Pu, Lei Yang, Xiaobao Xi'an University of Posts and Telecommunications School of Computer Science and Technology Xi'an China Rocket Force Engineering University School of Operational Support Xi'an China Xi'an University of Posts and Telecommunications Shaanxi Key Laboratory of Network Data Analysis and Intelligent Processing Xi'an China

ISBN: (纸本)9798350302219

To address the problem of insufficient representational capability and lack of online update of the Fully-convolutional Siamese Network (SiamFC) tracker in complex scenes, this paper proposes a siamese network visual tracking algorithm based on GCT attention and dual-template update mechanism. First, the feature extraction network is constructed by replacing AlexNet with the VGG16 network and SoftPool is used to replace the maximum pooling layer. Secondly, the attention module is added after the backbone network to enhance the network's ability to extract object features. Finally, a dual-template update mechanism is designed for response map fusion. Average Peak-to-Correlation Energy (APCE) is used to determine whether to update the dynamic templates, effectively improving the tracking robustness. The proposed algorithm is trained on the Got-10k dataset and tested on the OTB2015 and VOT2018 datasets. The experimental results show that, compared with SiamFC, the success rate and accuracy reach 0.663 and 0.891 on the OTB2015, which improve respectively 7.6% and 11.9%;On the VOT2018 dataset, the tracking accuracy, robustness and EAO are improved respectively by 2.9%, 29% and 14%. The proposed algorithm achieves high tracking accuracy in complex scenes and the tracking speed reaches 52.6 Fps, which meets the real-time tracking requirements. © 2023 IEEE.

关键词： Complex networks

Quiz-based Knowledge Tracing

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Shen, Shuanghong Chen, Enhong Xu, Bihan Liu, Qi Huang, Zhenya Zhu, Linbo Su, Yu Anhui Province Key Laboratory of Big Data Analysis and Application School of Data Science School of Computer Science and Techonology University of Science and Technology of China State Key Laboratory of Cognitive Intelligence Anhui Hefei230026 China Anhui Province Key Laboratory of Big Data Analysis and Application School of Data Science School of Computer Science and Techonology University of Science and Technology of China State Key Laboratory of Cognitive Intelligence Institute of Artificial Intelligence Hefei Comprehensive National Science Center Anhui Hefei230026 China School of Computer Science and Technology Hefei Normal University Institute of Artificial Intelligence Hefei Comprehensive National Science Center Anhui Hefei230601 China Institute of Artificial Intelligence Hefei Comprehensive National Science Center Anhui Hefei230026 China

Knowledge tracing (KT) aims to assess individuals’ evolving knowledge states according to their learning interactions with different exercises in online learning systems (OIS), which is critical in supporting decision-making for subsequent intelligent services, such as personalized learning source recommendation. Existing researchers have broadly studied KT and developed many effective methods. However, most of them assume that students’ historical interactions are uniformly distributed in a continuous sequence, ignoring the fact that actual interaction sequences are organized based on a series of quizzes with clear boundaries, where interactions within a quiz are consecutively completed, but interactions across different quizzes are discrete and may be spaced over days. In this paper, we present the Quiz-based Knowledge Tracing (QKT) model to monitor students’ knowledge states according to their quiz-based learning interactions. Specifically, as students’ interactions within a quiz are continuous and have the same or similar knowledge concepts, we design the adjacent gate followed by a global average pooling layer to capture the intra-quiz short-term knowledge influence. Then, as various quizzes tend to focus on different knowledge concepts, we respectively measure the inter-quiz knowledge substitution by the gated recurrent unit and the inter-quiz knowledge complementarity by the self-attentive encoder with a novel recency-aware attention mechanism. Finally, we integrate the inter-quiz long-term knowledge substitution and complementarity across different quizzes to output students’ evolving knowledge states. Extensive experimental results on three public real-world datasets demonstrate that QKT achieves state-of-the-art performance compared to existing methods. Further analyses confirm that QKT is promising in designing more effective quizzes. Copyright © 2023, The Authors. All rights reserved.

关键词： Students

Iteration Graph Network

学校读者我要写书评

暂无评论

Iteration Graph Network

2021 International Conference on Information, Cybernetics, and Computational Social Systems, ICCSS 2021

作者： Zhang, Wenchuan Ou, Weihua Niu, Shili Wang, Ruxin Zhu, Ziqi Ke, Shen Guizhou Normal University School of Big Data and Computer Science Guiyang China Yunnan University National Pilot School of Software Kunming China Wuhan University of Science and Technology School of Computer Science and Technology Wuhan China

ISBN: (纸本)9781665402453

Graph neural networks are widespreadly used in the field of graph data analysis and processing. Recent methods either reduce the spatial receptive field for low algorithm complexity, or greatly lose efficiency in order to realize attention mechanism. To tackle this issue, we propose Iteration Graph Network (IGN), which uses an iterative inversion method to aggregate node feature and the k-localized neighbor information of nodes. In the graph-based semi-supervised node classification task, our method surpasses the state-of-the-art method in the benchmark datasets and experiment conclusion show that our model outperforms graph attention networks (GAT) and is more than 3 times faster than graph attention networks, consumes more than 6 times less memory than GAT. Our code will be make publicly available. © 2021 IEEE.

关键词： Graph neural networks

Scalable Algorithm for Finding Balanced Subgraphs with Tolerance in Signed Networks

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Chen, Jingbang Peng, Richard Mang, Qiuyang Gao, Yu Zhou, Hangrui Ma, Chenhao David R. Cheriton School of Computer Science University of Waterloo Waterloo Canada Computer Science Department Carnegie Mellon University Pittsburgh United States School of Data Science The Chinese University of Hong Kong Shenzhen China Independent Beijing China Tsinghua University Beijing China

Signed networks, characterized by edges labeled as either positive or negative, offer nuanced insights into interaction dynamics beyond the capabilities of unsigned graphs. Central to this is the task of identifying the maximum balanced subgraph, crucial for applications like polarized community detection in social networks and portfolio analysis in finance. Traditional models, however, are limited by an assumption of perfect partitioning, which fails to mirror the complexities of real-world data. Addressing this gap, we introduce an innovative generalized balanced subgraph model that incorporates tolerance for imbalance. Our proposed region-based heuristic algorithm, tailored for this NP-hard problem, strikes a balance between low time complexity and high-quality outcomes. Comparative experiments validate its superior performance against leading solutions, delivering enhanced effectiveness (notably larger subgraph sizes) and efficiency (achieving up to 100× speedup) in both traditional and generalized contexts. Copyright © 2024, The Authors. All rights reserved.

关键词： Heuristic algorithms

Preventing Representational Rank Collapse in MPNNs by Splitting the Computational Graph

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Roth, Andreas Bause, Franka Kriege, Nils M. Liebig, Thomas Faculty of Computer Science TU Dortmund University Dortmund Germany Faculty of Computer Science University of Vienna Vienna Austria UniVie Doctoral School Computer Science University of Vienna Vienna Austria Research Network Data Science University of Vienna Vienna Austria Lamarr Institute for Machine Learning and Artificial Intelligence Dortmund Germany

The ability of message-passing neural networks (MPNNs) to fit complex functions over graphs is limited as most graph convolutions amplify the same signal across all feature channels, a phenomenon known as rank collapse, and oversmoothing as a special case. Most approaches to mitigate over-smoothing extend common message-passing schemes, e.g., the graph convolutional network, by utilizing residual connections, gating mechanisms, normalization, or regularization techniques. Our work contrarily proposes to directly tackle the cause of this issue by modifying the message-passing scheme and exchanging different types of messages using multi-relational graphs. We identify a sufficient condition to ensure linearly independent node representations. As one instantion, we show that operating on multiple directed acyclic graphs always satisfies our condition and propose to obtain these by defining a strict partial ordering of the nodes. We conduct comprehensive experiments that confirm the benefits of operating on multi-relational graphs to achieve more informative node representations. Copyright © 2024, The Authors. All rights reserved.

关键词： Convolution

Spiking transformer with experts mixture 24

学校读者我要写书评

暂无评论

Spiking transformer with experts mixture

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Zhaokun Zhou Yijie Lu Yanhao Jia Kaiwei Che Jun Niu Liwei Huang Xinyu Shi Yuesheng Zhu Guoqi Li Zhaofei Yu Li Yuan School of Electronic and Computer Engineering Shenzhen Graduate School Peking University and Peng Cheng Laboratory School of Electronic and Computer Engineering Shenzhen Graduate School Peking University College of Computing and Data Science Nanyang Technological University and Deep NeuroCognition Lab I2R and CFAR Agency for Science Technology and Research School of Computer Science Peking University and Peng Cheng Laboratory Institute for Artificial Intelligence Peking University and School of Computer Science Peking University Institute of Automation Chinese Academy of Sciences and Peng Cheng Laboratory

ISBN: (纸本)9798331314385

Spiking Neural Networks (SNNs) provide a sparse spike-driven mechanism which is believed to be critical for energy-efficient deep learning. Mixture-of-Experts (MoE), on the other side, aligns with the brain mechanism of distributed and sparse processing, resulting in an efficient way of enhancing model capacity and conditional computation. In this work, we consider how to incorporate SNNs' spike-driven and MoE's conditional computation into a unified framework. However, MoE uses softmax to get the dense conditional weights for each expert and TopK to hard-sparsify the network, which does not fit the properties of SNNs. To address this issue, we reformulate MoE in SNNs and introduce the Spiking Experts Mixture Mechanism (SEMM) from the perspective of sparse spiking activation. Both the experts and the router output spiking sequences, and their element-wise operation makes SEMM computation spike-driven and dynamic sparse-conditional. By developing SEMM into Spiking Transformer, the Experts Mixture Spiking Attention (EMSA) and the Experts Mixture Spiking Perceptron (EMSP) are proposed, which performs routing allocation for head-wise and channel-wise spiking experts, respectively. Experiments show that SEMM realizes sparse conditional computation and obtains a stable improvement on neuromorphic and static datasets with approximate computational overhead based on the Spiking Transformer baselines.

关键词：