检索结果-内蒙古大学图书馆

pFedCal: Lightweight Personalized Federated Learning with Adaptive Calibration Strategy

IEEE Transactions on Services Computing 2025年第03期18卷 1627-1640页

作者： Deng, Dongshang Wu, Xuangou Zhang, Tao Xiang, Chaocan Zhao, Wei Xu, Minrui Kang, Jiawen Han, Zhu Niyato, Dusit Anhui University of Technology School of Computer Science and Technology Anhui Province Key Laboratory of Digital Twin Technology in Metallurgical Industry Anhui Engineering Research Center for Intelligent Applications and Security of Industrial Internet Ma'anshan243002 China Beijing Jiaotong University School of Cyberspace Science and Technology Beijing100044 China Anhui Provincial Key Laboratory of Digital Twin Technology in Metallurgical Industry Anhui243002 China Chongqing University College of Computer Science Chongqing400044 China Nanyang Technological University School of Computer Science and Engineering 639798 Singapore Guangdong University of Technology School of Automation Guangzhou510062 China University of Houston Department of Electrical and Computer Engineering HoustonTX77004 United States Kyung Hee University Department of Computer Science and Engineering Seoul446-701 Korea Republic of

Federated learning (FL) is a promising artificial intelligence framework that enables clients to collectively train models with data privacy. However, in real-world scenarios, to construct practical FL frameworks, several challenges have to be addressed, including statistical heterogeneity, constrained resources, and fairness. Therefore, we first investigate an aggregation gap caused by statistical heterogeneity during local model initialization, which not only causes additional computational overhead for clients but also leads to the degradation of fairness. To bridge this gap, we propose pFedCal, a novel personalized federated learning with lightweight adaptive calibration strategy that performs calibration compensation through the prior knowledge of clients. Specifically, we introduce compensation for each client at the model initialization, with the compensation derived from the global gradient and the latest gradient bias. To enhance the calibration effect, we introduce a smoothing-based calibration strategy, and we design an adaptive calibration strategy. A representative example demonstrates that the proposed calibration and smoothing strategies improve fairness for clients. The theoretical analysis indicates that with an appropriate learning rate, pFedCal converges to a first-order stationary point for non-convex loss functions. Comprehensive experimental results show that pFedCal achieves faster convergence, higher accuracy, and improved fairness than the state-of-the-art methods. © 2008-2012 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

C-SPIN: Classification of Satellite Images for Prediction Network 14

C-SPIN: Classification of Satellite Images for Prediction Ne...

引用

14th International Conference on Cloud Computing, Data science and engineering, Confluence 2024

作者： Kumar, Sumit Mittal, Devanshu Nehra, Vibha Amity School of Engineering and Technology Department of Computer Science Engineering Noida India

ISBN: (纸本)9798350344837

Satellite image classification and prediction are crucial, with applications in domains ranging from environmental monitoring to urban planning, geological exploration, mapping, disaster response management, and agriculture. The paper provides a comprehensive analysis of current methods and techniques for object classification and prediction through satellite image data. The current work introduces the Classification of Satellite Images for Prediction Network (C-SPIN), which is meant to use both the spatial-geographical and temporal information inherent in satellite data. Image resolutions, compression, bit depth, sensor types, and spectral bands are among some of the variables, datasets, and factors considered in the research. Accurate object classification, detection, and prediction from satellite images are crucial for realising the full potential of the proposed model. The study gives insight on the strengths and limits of current models in various circumstances and highlights the potential capabilities of C-SPIN to enhance satellite image processing, image classification, and object prediction. © 2024 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Yoga pose classification: a CNN and MediaPipe inspired deep learning approach for real-world application

引用

Journal of Ambient Intelligence and Humanized Computing 2023年第12期14卷 16551-16562页

作者： Garg, Shubham Saxena, Aman Gupta, Richa Department of Computer Science and Engineering Amity School of Engineering and Technology Amity University Uttar Pradesh Noida India

Yoga is a centuries-old style of exercise followed by sports personnel, patients, and physiotherapist as their regime. A correct posture and technique are the key points in yoga to reap the maximum benefits. Hence, developing a model to classify yoga postures correctly is a recently emerging research topic. The paper presents a novel architecture that aims to classify various yoga poses. The proposed model estimates and classifies yoga poses into five broad categories with low latency. In the proposed architecture, the images are skeletonized before inputting into the model. The skeletonization process is done using the MediaPipe library for body keypoint detection. The paper compares the performance of various deep learning models with and without skeletonization. Different learning models showed the optimum result with the training of skeletonized images to the network. The comparison is drawn to establish the positive impact of skeletonization on the results obtained by various models. VGG16 achieves the highest validation accuracy on non-skeletonized images (95.6%), followed by InceptionV3, NASNetMobile, YogaConvo2d (proposed model) (89.9%), and lastly, InceptionResNetV2. In contrast, the proposed model YogaConvo2d using skeletonized images reports a validation accuracy of 99.62%, followed by VGG16, InceptionResNetV2, NASNetMobile, and InceptionV3. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

关键词： computer vision

来源：评论

学校读者我要写书评

暂无评论

circ2CBA: prediction of circRNA-RBP binding sites combining deep learning and attention mechanism

引用

Frontiers of computer science 2023年第5期17卷 217-225页

作者： Yajing GUO Xiujuan LEI Lian LIU Yi PAN School of Computer Science Shaanxi Normal UniversityXi’an 710119China Faculty of Computer Science and Control Engineering Shenzhen Institute of Advanced TechnologyChinese Academy of SciencesShenzhen 518055China

Circular RNAs(circRNAs)are RNAs with closed circular structure involved in many biological processes by key interactions with RNA binding proteins(RBPs).Existing methods for predicting these interactions have limitations in feature *** view of this,we propose a method named circ2CBA,which uses only sequence information of circRNAs to predict circRNA-RBP binding *** have constructed a data set which includes eight ***,circ2CBA encodes circRNA sequences using the one-hot ***,a two-layer convolutional neural network(CNN)is used to initially extract the *** CNN,circ2CBA uses a layer of bidirectional long and short-term memory network(BiLSTM)and the self-attention mechanism to learn the *** AUC value of circ2CBA reaches *** of circ2CBA with other three methods on our data set and an ablation experiment confirm that circ2CBA is an effective method to predict the binding sites between circRNAs and RBPs.

关键词： circRNAs RBPs CNN BiLSTM self-attention mechanism

来源：评论

学校读者我要写书评

暂无评论

A Comprehensive Exploration of Real-Time 3-D View Reconstruction Methods

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第12期5卷 5915-5927页

作者： Agrawal, Arya Sharma, Teena Verma, Nishchal K. Vellore Institute of Technology School of Computer Science and Engineering Vellore632014 India Indian Institute of Technology Mehta Family School of Data Science and Artificial Intelligence Guwahati781039 India Indian Institute of Technology Kanpur Department of Electrical Engineering 208016 India

Real-time 3-D view reconstruction in an unfamiliar environment poses complexity for various applications due to varying conditions such as occlusion, latency, precision, etc. This article thoroughly examines and tests contemporary methodologies addressing challenges in 3-D view reconstruction. The methods being explored in this article are categorized into volumetric and mesh, generative adversarial network based, and open source library based methods. The exploration of these methods undergoes detailed discussions, encompassing methods, advantages, limitations, and empirical results. The real-time testing of each method is done on benchmarked datasets, including ShapeNet, Pascal 3D+, Pix3D, etc. The narrative highlights the crucial role of 3-D view reconstruction in domains such as robotics, virtual and augmented reality, medical imaging, cultural heritage preservation, etc. The article also anticipates future scopes by exploring generative models, unsupervised learning, and advanced sensor fusion to increase the robustness of the algorithms. © 2020 IEEE.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

Masked Generative Light Field Prompting for Pixel-Level Structure Segmentations

引用

Research 2024年第4期2024卷 533-544页

作者： Mianzhao Wang Fan Shi Xu Cheng Shengyong Chen The Engineering Research Center of Learning-Based Intelligent System(Ministry of Education) Tianjin University of TechnologyTianjin 300384China Key Laboratory of Computer Vision and System(Ministry of Education) Tianjin University of TechnologyTianjin 300384China School of Computer Science and Engineering Tianjin University of TechnologyTianjin 300384China

Pixel-level structure segmentations have attracted considerable attention,playing a crucial role in autonomous driving within the metaverse and enhancing comprehension in light field-based machine ***,current light field modeling methods fail to integrate appearance and geometric structural information into a coherent semantic space,thereby limiting the capability of light field transmission for visual *** this paper,we propose a general light field modeling method for pixel-level structure segmentation,comprising a generative light field prompting encoder(LF-GPE)and a prompt-based masked light field pretraining(LF-PMP)*** LF-GPE,serving as a light field backbone,can extract both appearance and geometric structural cues *** aligns these features into a unified visual space,facilitating semantic ***,our LF-PMP,during the pretraining phase,integrates a mixed light field and a multi-view light field *** prioritizes considering the geometric structural properties of the light field,enabling the light field backbone to accumulate a wealth of prior *** evaluate our pretrained LF-GPE on two downstream tasks:light field salient object detection and semantic *** results demonstrate that LF-GPE can effectively learn high-quality light field features and achieve highly competitive performance in pixel-level segmentation tasks.

关键词： prompt backbone integrate

来源：评论

学校读者我要写书评

暂无评论

Solving Multi-Objective Vehicle Routing Problems with Time Windows: A Decomposition-Based Multiform Optimization Approach

引用

Tsinghua science and technology 2024年第2期29卷 305-324页

作者： Yiqiao Cai Zifan Lin Meiqin Cheng Peizhong Liu Ying Zhou Xiamen Key Laboratory of Data Security and Blockchain Technology College of Computer Science and TechnologyHuaqiao UniversityXiamen361021China College of Engineering Huaqiao UniversityQuanzhou362000China School of Computer Sciences Shenzhen Institute of Information TechnologyShenzhen518000China

In solving multi-objective vehicle routing problems with time windows (MOVRPTW),most existing algorithms focus on the optimization of a single problem formulation. However,little effort has been devoted to exploiting valuable knowledge from the alternate formulations of MOVRPTW for better optimization performance. Aiming at this insufficiency,this study proposes a decomposition-based multi-objective multiform evolutionary algorithm (MMFEA/D),which performs the evolutionary search on multiple alternate formulations of MOVRPTW simultaneously to complement each other. In particular,the main characteristics of MMFEA/D are three folds. First,a multiform construction (MFC) strategy is adopted to construct multiple alternate formulations,each of which is formulated by grouping several adjacent subproblems based on the decomposition of MOVRPTW. Second,a transfer reproduction (TFR) mechanism is designed to generate offspring for each formulation via transferring promising solutions from other formulations,making that the useful traits captured from different formulations can be shared and leveraged to guide the evolutionary search. Third,an adaptive local search (ALS) strategy is developed to invest search effort on different alternate formulations as per their usefulness for MOVRPTW,thus facilitating the efficient allocation of computational resources. Experimental studies have demonstrated the superior performance of MMFEA/D on the classical Solomon instances and the real-world instances.

关键词： MOVRPTW multiform optimization multiform construction knowledge transfer adaptive local search

来源：评论

学校读者我要写书评

暂无评论

SEAformer: frequency domain decomposition transformer with signal enhanced for long-term wind power forecasting

引用

Neural Computing and Applications 2024年第33期36卷 20883-20906页

作者： Yan, Leiming Wu, Siqi Li, Shaopeng Chen, Xianyi School of Computer Science Nanjing University of Information Science and Technology Nanjing China Engineering Research Center of Digital Forensics Ministry of Education Nanjing University of Information Science and Technology Nanjing China

Accurate and reliable wind power forecasting is of great importance for stable grid operation and advanced dispatch planning. Due to the complex, non-stationary, and highly volatile nature of wind power data, Transformer-based methods find it difficult to capture long-term trend features and incur high computational costs. To address these challenging problems, we propose a frequency domain decomposition Transformer architecture with signal enhanced attention mechanism (SEAformer). Firstly, we devise a frequency domain-based trend decomposition structure that enables the Transformer to extract more effective long-term trend features, thereby further improving the long-term prediction accuracy of the model. Secondly, in response to the large fluctuations and instability of wind power data, we design an internal signal enhanced substructure combined with an attention mechanism in the Transformer, which filters out high-frequency noise signals and reduces the computational cost of the Transformer. We conduct extensive experiments on the benchmark dataset, the experimental analysis demonstrates that SEAformer outperforms the baseline methods (Transformer-based, MLP-based, and Traditional methods) in both multivariate and univariate prediction tasks, exhibiting the best prediction performance. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Wind forecasting

来源：评论

学校读者我要写书评

暂无评论

Efficient Adaptation of Pre-trained Vision Transformer via Householder Transformation 38

Efficient Adaptation of Pre-trained Vision Transformer via H...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Dong, Wei Sun, Yuan Yang, Yiting Zhang, Xing Lin, Zhijun Yan, Qingsen Zhang, Haokui Wang, Peng Yang, Yang Shen, Hengtao College of Information and Control Engineering Xi'an University of Architecture and Technology China School of Computer Science Northwestern Polytechnical University China School of Computer Science and Engineering University of Electronic Science and Technology of China China School of Computer Science and Technology Tongji University China

A common strategy for Parameter-Efficient Fine-Tuning (PEFT) of pre-trained Vision Transformers (ViTs) involves adapting the model to downstream tasks by learning a low-rank adaptation matrix. This matrix is decomposed into a product of down-projection and up-projection matrices, with the bottleneck dimensionality being crucial for reducing the number of learnable parameters, as exemplified by prevalent methods like LoRA and Adapter. However, these low-rank strategies typically employ a fixed bottleneck dimensionality, which limits their flexibility in handling layer-wise variations. To address this limitation, we propose a novel PEFT approach inspired by Singular Value Decomposition (SVD) for representing the adaptation matrix. SVD decomposes a matrix into the product of a left unitary matrix, a diagonal matrix of scaling values, and a right unitary matrix. We utilize Householder transformations to construct orthogonal matrices that efficiently mimic the unitary matrices, requiring only a vector. The diagonal values are learned in a layer-wise manner, allowing them to flexibly capture the unique properties of each layer. This approach enables the generation of adaptation matrices with varying ranks across different layers, providing greater flexibility in adapting pre-trained models. Experiments on standard downstream vision tasks demonstrate that our method achieves promising fine-tuning performance. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Probabilistic Multi-Product Trading of Hydro Power Plants in Sequential Intraday and Frequency-Regulation Markets

IEEE Transactions on Energy Markets, Policy and Regulation

引用

IEEE Transactions on Energy Markets, Policy and Regulation 2024年第4期2卷 449-464页

作者： Mohammadi, Saeed Khodadadi, Abolfazl Shinde, Priyanka Blom, Evelin Hesamzadeh, Mohammad Reza Soder, Lennart KTH Royal Institute of Technology School of Electrical Engineering and Computer Science Stockholm100 44 Sweden

With the increasing integration of power plants into the frequency-regulation markets, the importance of optimal trading has grown substantially. This paper conducts an in-depth analysis of their optimal trading behavior in sequential day-ahead, intraday, and frequency-regulation markets. We introduce a probabilistic multi-product optimization model, derived through a series of transformation techniques. Additionally, we present two reformulations that re-frame the problem as a mixed-integer linear programming problem with uncertain parameters. Various aspects of the model are thoroughly examined to observe the optimal multi-product trading behavior of hydro power plant assets, along with numerous case studies. Leveraging historical data from Nordic electricity markets, we construct realistic scenarios for the uncertain parameters. Furthermore, we then proposed an algorithm based on the No-U-Turn sampler to provide probability distribution functions of cleared prices in frequency-regulation and day-ahead markets. These distribution functions offer valuable statistical insights into temporal price risks for informed multi-product optimal-trading decisions. © 2024 IEEE.

关键词： Power markets

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：