检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Sohn, Jy-Yong Kwon, Dohyun An, Seoyeon Lee, Kangwook Department of Statistics and Data Science Yonsei University Korea Republic of Department of Mathematics University of Seoul Korea Republic of Center for AI and Natural Sciences Korea Institute for Advanced Study Korea Republic of Department of Electrical and Computer Engineering University of Wisconsin-Madison WI United States

Fine-tuning large pre-trained models is a common practice in machine learning applications, yet its mathematical analysis remains largely unexplored. In this paper, we study fine-tuning through the lens of memorization capacity. Our new measure, the Fine-Tuning Capacity (FTC), is defined as the maximum number of samples a neural network can fine-tune, or equivalently, as the minimum number of neurons (m) needed to arbitrarily change N labels among K samples considered in the fine-tuning process. In essence, FTC extends the memorization capacity concept to the fine-tuning scenario. We analyze FTC for the additive fine-tuning scenario where the fine-tuned network is defined as the summation of the frozen pre-trained network f and a neural network g (with m neurons) designed for fine-tuning. When g is a ReLU network with either 2 or 3 layers, we obtain tight upper and lower bounds on FTC;we show that N samples can be fine-tuned with m = Θ(N) neurons for 2-layer networks, and with m = Θ(√N) neurons for 3-layer networks, no matter how large K is. Our results recover the known memorization capacity results when N = K as a special case. Copyright © 2024, The Authors. All rights reserved.

关键词： Neurons

来源：评论

学校读者我要写书评

暂无评论

Privacy Preservation for Cloud-Edge-Collaborative Energy Management System Using Post-Quantum Homomorphic Encryption

引用

IEEE Transactions on Smart Grid 2025年第4期16卷 3282-3294页

作者： Jiang, Cheng Li, Xue Du, Dajun Qian, Hong Wen, Mi Wu, Lei Findeisen, Rolf Shanghai University of Electric Power Shanghai Key Laboratory of Power Station Automation Technology College of Automation Engineering Shanghai200090 China Shanghai University Shanghai Key Laboratory of Power Station Automation Technology School of Mechatronical Engineering and Automation Shanghai200072 China Shanghai University of Electric Power College of Computer Science and Technology Shanghai201306 China Advanced Cryptography and System Security Key Laboratory of Sichuan Province Chengdu610225 China Stevens Institute of Technology Department of Electrical and Computer Engineering HobokenNJ07030 United States TU Darmstadt Control and Cyber-Physical Systems Laboratory Darmstadt64227 Germany

Cloud-based energy management systems (EMS) in smart grids face privacy challenges, as existing methods based on traditional homomorphic encryption support limited operations and are vulnerable to quantum attacks. We propose a privacy preservation method for smart grid energy management system, leveraging edge-assisted computing and ring learning with errors (RLWE) homomorphic encryption. Our approach enables both addition and multiplication operations on encrypted smart grid data while defending against quantum threats. We demonstrate the methods security under the RLWE hardness assumption and validate its effectiveness through an experimental platform simulating smart grid environments. This work addresses computational limitations in current privacy-preserving techniques, enhancing data security for evolving smart grid technologies © 2010-2012 IEEE.

关键词： Power management (telecommunication)

来源：评论

学校读者我要写书评

暂无评论

TREE QUASI-SEPARABLE MATRICES: A SIMULTANEOUS GENERALIZATION OF SEQUENTIALLY AND HIERARCHICALLY SEMI-SEPARABLE REPRESENTATIONS

arXiv

引用

arXiv 2024年

作者： Govindarajan, Nithin Chandrasekaran, Shivkumar Dewilde, Patrick KU Leuven Leuven Belgium University of California Santa BarbaraCA United States Electrical Engineering Mathematics and Computer Science TU Delft Delft Netherlands Institute of Advanced Study Technical University Munich Germany

We present a unification and generalization of what is known in the literature as sequentially and hierarchically semi-separable (SSS and HSS) representations for matrices. Describing rank-structured representations of (inverses of) sparse matrices whose adjacency graph is a tree, it is shown that these so-called tree quasi-separable (TQS) matrices inherit all the favorable algebraic properties of SSS and HSS under addition, products, and inversion. To arrive at these properties, we prove a key result that characterizes the conversion of any dense matrix into a TQS representation. Here, we specifically show through an explicit construction procedure that the generator sizes are dictated by the ranks of certain Hankel blocks of the matrix. Analogous to SSS and HSS, TQS matrices admit fast matrix-vector products and direct solvers provided the generator sizes are small. A sketch of the associated algorithms is *** Codes 15A23 © 2024, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Control of Magnetization Precession Frequency Using Current Density of Spin-Orbit Torque Applied to Local Area

Control of Magnetization Precession Frequency Using Current ...

引用

2023 IEEE International Magnetic Conference - Short Papers, INTERMAG Short Papers 2023

作者： Koda, Tetsunori Nakagawa, Ao Muroga, Sho Endo, Yasushi Oshima College National Institute of Technology Department of Electronic Mechanical Engineering Suo-Oshima742-2193 Japan Akita University Graduate School of Engineering Science Department of Mathematical Science and Electrical-Electronic-Computer Engineering Akita010-8502 Japan Department of Electrical Engineering Graduate School of Engineering Tohoku University Sendai980-8579 Japan Center for Spintronics Research Network Tohoku University Sendai980-8577 Japan Organization for Advanced Studies Tohoku University Sendai980-8577 Japan

ISBN: (纸本)9798350338362

We numerically study current density dependence of magnetization dynamics in a local area induced by the spin-orbit torque as a source of spin waves. The precession frequency of magnetization dynamics shows large change at a critical current density, which controls the wave length of spin waves. The origin of the change is due to the increase of exchange field caused by the magnetization torsion at the interface between SOT-applied and non-applied regions. Our new finding gives a way to manipulate spin wave's properties by controlling the magnetization direction in a local area. © 2023 IEEE.

关键词： Spin waves

来源：评论

学校读者我要写书评

暂无评论

Enhancing Power Quality Event Classification with AI Transformer Models

Enhancing Power Quality Event Classification with AI Transfo...

引用

IEEE General Meeting Power& Energy Society

作者： Ahmad M. Saber Amr Youssef Davor Svetinovic Hatem Zeineldin Deepa Kundur Ehab El-Saadany EECS Department Advanced Power and Energy Center Khalifa University Abu Dhabi UAE Department of Electrical and Computer Engineering University of Toronto Toronto Canada Concordia Institute for Information Systems Engineering Montreal Canada

ISBN: (数字)9798350381832

ISBN: (纸本)9798350381849

Recently, there has been a growing interest in utilizing machine learning for accurate classification of power quality events (PQEs). However, most of these studies are performed assuming an ideal situation, while in reality, we can have measurement noise, DC offset, and variations in the voltage signal’s amplitude and frequency. Building on the prior PQE classification works using deep learning, this paper proposes a deep-learning framework that leverages attention-enabled Transformers as a tool to accurately classify PQEs under the aforementioned considerations. The proposed framework can operate directly on the voltage signals with no need for a separate feature extraction or calculation phase. Our results show that the proposed framework outperforms recently proposed learning-based techniques. It can accurately classify PQEs under the aforementioned conditions with an accuracy varying between 99.81%–91.43% depending on the signal-to-noise ratio, DC offsets, and variations in the signal amplitude and frequency.

关键词： Voltage measurement Noise Power quality Transformers Harmonic analysis Frequency measurement Numerical models Noise measurement Transient analysis Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

A Review of Clinical Applications for Super-resolution Ultrasound Localization Microscopy

引用

Current Medical Science 2022年第1期42卷 1-16页

作者： Hui-ming YI Matthew RLowerison Peng-fei SONG Wei ZHANG eDpartment of Medical Ultrasound Tongji HospitalTongji Medical CollegeHuazhong University of Science and TechnologyWuhan 430030China Eckman Institute for Advanced Science and Technology University of Illinois at Urbana-ChampaignUrbana 61801USA Department of Electrical and Computer Engineering University of Illinois at Urbana-ChampaignUrbana 61801USA

Microvascular structure and hemodynamics are important indicators for the diagnosis and assessment of many diseases and *** structural and functional imaging of tissue microvasculature in vivo is a clinically significant objective for the development of many imaging ***-enhanced ultrasound(CEUS)is a popular clinical tool for characterizing tissue microvasculature,due to the moderate cost,wide accessibility,and absence of ionizing radiation of ultrasound.

关键词： contrast-enhanced ultrasound super-resolution ultrasound microvascular imaging microbubbles

来源：评论

学校读者我要写书评

暂无评论

Guest Editorial: Special issue on computational methods and artificial intelligence applications in low-carbon energy systems

引用

IET Renewable Power Generation 2024年第3期18卷 303-305页

作者： Wang, Yishen Zhou, Fei Guerrero, Josep M. Baker, Kyri Chen, Yize Wang, Hao Xu, Bolun Xu, Qianwen Zhu, Hong Agwan, Utkarsha State Grid Laboratory of Grid Advanced Computing and Applications State Grid Smart Grid Research Institute Co. Ltd. Beijing China Department of Energy Technology Aalborg University Aalborg Denmark Department of Civil Environmental and Architectural Engineering University of Colorado-Boulder BoulderCO United States Guangzhou China Department of Data Science and AI Monash University ClaytonVIC Australia Department of Earth and Environmental Engineering Columbia University New YorkNY United States Electric Power and Energy Systems Division KTH Royal Institute of Technology Stockholm Sweden State Grid Jiangsu Electric Power Co. Ltd. Nanjing China Department of Electrical Engineering and Computer Sciences University of California - Berkeley BerkeleyCA United States

来源：评论

学校读者我要写书评

暂无评论

Large Scale Model Enabled Semantic Communications Based on Robust Knowledge Distillation

Large Scale Model Enabled Semantic Communications Based on R...

引用

2024 IEEE Global Communications Conference, GLOBECOM 2024

作者： Ding, Kuiyuan Liu, Fangfang Yang, Yang Chen, Mingzhe Guo, Caili Beijing University of Posts and Telecommunications Beijing Key Laboratory of Network System Architecture and Convergence School of Information and Communication Engineering Beijing100876 China University of Miami Department of Electrical and Computer Engineering The Institute for Data Science and Computing Coral GablesFL33146 United States Beijing University of Posts and Telecommunications Beijing Laboratory of Advanced Information Networks School of Information and Communication Engineering Beijing100876 China

ISBN: (纸本)9798350351255

Large scale artificial intelligence (AI) models possess excellent capabilities in semantic representation and understanding, making them particularly well-suited for semantic encoding and decoding. However, the substantial scale of these AI models imposes unacceptable computational resources and communication delays. To address this issue, we propose a semantic communication scheme based on robust knowledge distillation (RKD-SC) for large scale model enabled semantic communications. In the considered system, a transmitter extracts the features of the source image for robust transmission and accurate image classification at the receiver. To effectively utilize the superior capability of large scale model while make the cost affordable, we first transfer knowledge from a large scale model to a smaller scale model to serve as the semantic encoder. Then, to enhance the robustness of the system against channel noise, we propose a channel-aware autoencoder (CAA) based on the Transformer architecture. Experimental results show that the encoder of proposed RKD-SC system can achieve over 93.3% of the performance of a large scale model while compressing 96.67% number of parameters. Code: https://***/echojayne/RKD-SC. © 2024 IEEE.

关键词： Channel coding

来源：评论

学校读者我要写书评

暂无评论

Revealing Non-Hermitian Band Structure of Photonic Floquet Media

Revealing Non-Hermitian Band Structure of Photonic Floquet M...

引用

Frontiers in Optics, FiO 2022

作者： Park, Jagang Cho, Hyukjoon Lee, Seojoo Lee, Kyungmin Lee, Kanghee Park, Hee Chul Ryu, Jung-Wan Park, Namkyoo Jeon, Sanggeun Min, Bumki Department of Mechanical Engineering Korea Advanced Institute of Science and Technology Daejeon34141 Korea Republic of Department of Physics Korea Advanced Institute of Science and Technology Daejeon34141 Korea Republic of Center for Theoretical Physics of Complex Systems Institute of Basic Science Daejeon34126 Korea Republic of Department of Electrical and Computer Engineering Seoul National University Seoul08826 Korea Republic of School of Electrical Engineering Korea University Seoul02841 Korea Republic of

We experimentally reveal the unique non-Hermitian band structures of a photonic Floquet medium emulated in the microwave regime. We observe the presence of the momentum gaps and the subharmonic parametric oscillations...

ISBN: (纸本)9781557528209

关键词：

来源：评论

学校读者我要写书评

暂无评论

Deduplication of Textual Data by NLP Approaches

Deduplication of Textual Data by NLP Approaches

引用

IEEE Conference on Vehicular Technology (VTC)

作者： Kiana Ghassabi Peyman Pahlevani Daniel E. Lucani Department of Computer Science Institute for Advanced Studies in Basic Sciences (IASBS) Zanjan Iran Department of Electronic Systems Aalborg University Copenhagen Denmark Department of Electrical and Computer Engineering Aarhus University Aarhus Denmark

With the increasing amount of digital data, data deduplication has become an increasingly popular method for reducing data in large-scale storage systems. Generalized deduplication is an alternative technique for reducing the cost of data storage by identifying similar data chunks. This paper proposes TL-GD, a method for improving cloud storage efficiency using generalized deduplication focusing on textual datasets. The core concept of this study is to develop an efficient deduplication system that combines an alternative technique for splitting data into smaller pieces and a new approach for transforming data pieces into bases and deviations. The performance of the system has been validated using two real-world datasets. We also compare the results to state-of-the-art deduplication methods. Our evaluation results show that TL-GD achieves nearly 67% lossless compression for textual navigation instructions datasets, which is a 25% improvement on average compared to existing deduplication techniques.

关键词：

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：