检索结果-内蒙古大学图书馆

On the generalization of learning algorithms that do not converge

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Chandramoorthy, Nisha Loukas, Andreas Gatmiry, Khashayar Jegelka, Stefanie Institute for Data Systems and Society Massachusetts Institute of Technology CambridgeMA02139 United States Prescient Design Genentech Roche United States Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA02139 United States

Generalization analyses of deep learning typically assume that the training converges to a fixed point. But, recent results indicate that in practice, the weights of deep neural networks optimized with stochastic gradient descent often oscillate indefinitely. To reduce this discrepancy between theory and practice, this paper focuses on the generalization of neural networks whose training dynamics do not necessarily converge to fixed points. Our main contribution is to propose a notion of statistical algorithmic stability (SAS) that extends classical algorithmic stability to non-convergent algorithms and to study its connection to generalization. This ergodic-theoretic approach leads to new insights when compared to the traditional optimization and learning theory perspectives. We prove that the stability of the time-asymptotic behavior of a learning algorithm relates to its generalization and empirically demonstrate how loss dynamics can provide clues to generalization performance. Our findings provide evidence that networks that "train stably generalize better" even when the training continues indefinitely and the weights do not converge. Copyright © 2022, The Authors. All rights reserved.

关键词： Learning algorithms

Inferring Coflow Size Based on Broad Learning System in Data Center Network 20

学校读者我要写书评

暂无评论

Inferring Coflow Size Based on Broad Learning System in Data...

20th IEEE International Conference on Communication Technology, ICCT 2020

作者： Ye, Jin Ma, Liangdi Pang, Chengjie Xiao, Qingyu Jiang, Wenchao Guangxi University School of Computer Electronics and Information Nanning China The University of Hongkong Department of Electrical and Electronic Engineering China Singapore University of Technology and Design Pillar of Information Systems Technology and Design Singapore Singapore

ISBN: (纸本)9781728181417

In recent years, coflow size-based scheduling has become a hotspot in data center networks since it can effectively optimize the coflow completion time (CCT). However, it is a challenging work to predict the coflow size accurately and rapidly with limited information. This paper proposes a Broad Learning System based coflow size inferring model, which employs Multi -RELIEF method to extract features and trains BLS model on Facebook logs with incomplete information. Experiment results show that the accuracy and sensitivity of BLS-based inferring model are higher than the ELM and XGBoost methods by 31.6% when we use only half of coflow information, and the prediction time remains below 6ms that meets the online processing requirement of most data center networks. Besides, the simulated scheduling results indicate that the CCT of the BLS-based coflow scheduler outperforms the FIFO, Aalo and MCS schedulers at most time. © 2020 IEEE.

关键词： Learning systems

Achievable Diversity Order of HARQ-Aided Downlink NOMA systems

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Shi, Zheng Zhang, Chenmeng Fu, Yaru Wang, Hong Yang, Guanghua Ma, Shaodan The Institute of Physical Internet The School of Intelligent Systems Science and Engineering Jinan University Zhuhai519070 China The School of Telecommunications Engineering Xidian University Xi’an710000 China The Department of Information Systems Technology and Design Singapore University of Technology and Design Singapore The School of Communication and Information Engineering Nanjing University of Posts and Telecommunications Nanjing210003 China The National Mobile Communications Research Laboratory Southeast University Nanjing210096 China The State Key Laboratory of Internet of Things for Smart City The Department of Electrical and Computer Engineering University of Macau China

The combination between non-orthogonal multiple access (NOMA) and hybrid automatic repeat request (HARQ) is capable of realizing ultra-reliability, high throughput and many concurrent connections particularly for emerging communication systems. This paper focuses on characterizing the asymptotic scaling law of the outage probability of HARQ-aided NOMA systems with respect to the transmit power, i.e., diversity order. The analysis of diversity order is carried out for three basic types of HARQ-aided downlink NOMA systems, including Type I HARQ, HARQ with chase combining (HARQ-CC) and HARQ with incremental redundancy (HARQ-IR). The diversity orders of three HARQ-aided downlink NOMA systems are derived in closed-form, where an integration domain partition trick is developed to obtain the bounds of the outage probability specially for HARQ-CC and HARQ-IR-aided NOMA systems. The analytical results show that the diversity order is a decreasing step function of transmission rate, and full time diversity can only be achieved under a sufficiently low transmission rate. It is also revealed that HARQ-IR-aided NOMA systems have the largest diversity order, followed by HARQ-CC-aided and then Type I HARQ-aided NOMA systems. Additionally, the users’ diversity orders follow a descending order according to their respective average channel gains. Furthermore, we expand discussions on the cases of power-efficient transmissions and imperfect channel state information (CSI). Monte Carlo simulations finally confirm our analysis. © 2022, CC BY.

关键词： Monte Carlo methods

Hardware implementation of memristor-based artificial neural networks

学校读者我要写书评

暂无评论

Nature Communications 2024年第1期15卷 1974页

作者： Aguirre, Fernando Sebastian, Abu Le Gallo, Manuel Song, Wenhao Wang, Tong Yang, J. Joshua Lu, Wei Chang, Meng-Fan Ielmini, Daniele Yang, Yuchao Mehonic, Adnan Kenyon, Anthony Villena, Marco A. Roldán, Juan B. Wu, Yuting Hsu, Hung-Hsi Raghavan, Nagarajan Suñé, Jordi Miranda, Enrique Eltawil, Ahmed Setti, Gianluca Smagulova, Kamilya Salama, Khaled N. Krestinskaya, Olga Yan, Xiaobing Ang, Kah-Wee Jain, Samarth Li, Sifan Alharbi, Osamah Pazos, Sebastian Lanza, Mario Physical Science and Engineering Division King Abdullah University of Science and Technology (KAUST) Thuwal 23955-6900 Saudi Arabia Departament d’Enginyeria Electrònica Universitat Autònoma de Barcelona (UAB) Barcelona 08193 Spain IBM Research – Zurich Rüschlikon Switzerland Department of Electrical and Computer Engineering University of Southern California (USC) Los Angeles 90089 CA United States Department of Electrical Engineering and Computer Science University of Michigan Ann Arbor 48109 MI United States Department of Electrical Engineering National Tsing Hua University Hsinchu 30013 Taiwan Dipartimento di Elettronica Informazione e Bioingegneria Politecnico di Milano and IUNET Piazza L. da Vinci 32 Milano 20133 Italy School of Electronic and Computer Engineering Peking University Shenzhen China Department of Electronic and Electrical Engineering University College London (UCL) Torrington Place WC1E 7JE London United Kingdom Departamento de Electrónica y Tecnología de Computadores Facultad de Ciencias Universidad de Granada Avenida Fuentenueva s/n Granada 18071 Spain Engineering Product Development (EPD) Pillar Singapore University of Technology & amp Computer Electrical and Mathematical Sciences and Engineering Division King Abdullah University of Science and Technology (KAUST) Thuwal 23955-6900 Saudi Arabia Key Laboratory of Brain-Like Neuromorphic Devices and Systems of Hebei Province Hebei University Baoding 071002 China Department of Electrical and Computer Engineering College of Design and Engineering National University of Singapore (NUS) Singapore Singapore

Artificial Intelligence (AI) is currently experiencing a bloom driven by deep learning (DL) techniques, which rely on networks of connected simple computing units operating in parallel. The low communication bandwidth between memory and processing units in conventional von Neumann machines does not support the requirements of emerging applications that rely extensively on large sets of data. More recent computing paradigms, such as high parallelization and near-memory computing, help alleviate the data communication bottleneck to some extent, but paradigm- shifting concepts are required. Memristors, a novel beyond-complementary metal-oxide-semiconductor (CMOS) technology, are a promising choice for memory devices due to their unique intrinsic device-level properties, enabling both storing and computing with a small, massively-parallel footprint at low power. Theoretically, this directly translates to a major boost in energy efficiency and computational throughput, but various practical challenges remain. In this work we review the latest efforts for achieving hardware-based memristive artificial neural networks (ANNs), describing with detail the working principia of each block and the different design alternatives with their own advantages and disadvantages, as well as the tools required for accurate estimation of performance metrics. Ultimately, we aim to provide a comprehensive protocol of the materials and methods involved in memristive neural networks to those aiming to start working in this field and the experts looking for a holistic approach. © The Author(s) 2024.

关键词：

CAN Signal Processing for EMI Reduction Cooperating with Switched-Mode Power Supply 12

学校读者我要写书评

暂无评论

CAN Signal Processing for EMI Reduction Cooperating with Swi...

12th Annual IEEE Energy Conversion Congress and Exposition, ECCE 2020

作者： Shirai, Ryo Wada, Keiji Shimizu, Toshihisa Tokyo Metropolitan University Graduate School of Systems Design Department of Electrical Engineering and Computer Science 1-1 Minami-Osawa Hachioji-shi Tokyo192-0397 Japan

ISBN: (纸本)9781728158266

Electromagnetic interference (EMI) on wired communication systems caused by a switched-mode power supply is often remarkable in practice. The EMI undermines the performance of the communication and robust operation of the system. In this paper, a test system where a buck converter and a controller area network (CAN) communication system are integrated enables to measure noise waveforms and analyze the EMI mechanism in the time-domain. The simulation model of the system also performs verification of the mechanism. In the system, the DC bus cables connected to the converter are closely adjacent to the communication cables and they have electromagnetic couplings. Hence, when the converter generates a switching noise in the DC bus, an induced noise happens in the communication cable, which results in communication failure. Based on the EMI mechanism, a noise elimination method is developed and implemented to the CAN of the test system to prevent communication failure. The proposed method is feasible in the converter controller and it does not require additional hardware components to reduce the communication failure. Through experimental verification, the proposed method enables to reduce communication failure. © 2020 IEEE.

关键词： DC-DC converters

VISUALTTS: TTS with ACCURATE LIP-SPEECH SYNCHRONIZATION for AUTOMATIC VOICE over

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Lu, Junchen Sisman, Berrak Liu, Rui Zhang, Mingyang Li, Haizhou Department of Electrical and Computer Engineering National University of Singapore Singapore Information Systems Technology and Design Singapore University of Technology and Design Singapore The Chinese University of Hong Kong Shenzhen China

In this paper, we formulate a novel task to synthesize speech in sync with a silent pre-recorded video, denoted as automatic voice over (AVO). Unlike traditional speech synthesis, AVO seeks to generate not only human-sounding speech, but also perfect lip-speech synchronization. A natural solution to AVO is to condition the speech rendering on the temporal progression of lip sequence in the video. We propose a novel text-to-speech model that is conditioned on visual input, named VisualTTS, for accurate lip-speech synchronization. The proposed VisualTTS adopts two novel mechanisms that are 1) textual-visual attention, and 2) visual fusion strategy during acoustic decoding, which both contribute to forming accurate alignment between the input text content and lip motion in input lip sequence. Experimental results show that VisualTTS achieves accurate lip-speech synchronization and outperforms all baseline systems. Copyright © 2021, The Authors. All rights reserved.

关键词： Speech synthesis

Implementation of Grover's algorithm on the IBMQ: Constraints and error assessment

学校读者我要写书评

暂无评论

Implementation of Grover's algorithm on the IBMQ: Constraint...

Photonics for Quantum 2021

作者： Eldridi, Asma Gueddana, Amor Lakshminarayanan, Vasudevan University of Carthage Department of Electrical Engineering National Engineering School of Carthage ENICarthage Tunis2035 Tunisia University of Carthage Higher School of Communication of Tunis Sup'Com Green and Smart Communication Systems Lab Gres'Com Ariana2083 Tunisia University of Waterloo School of Optometry and Vision Science Theoretical and Experimental Epistemology Lab TEEL 200 University Avenue West WaterlooONN2l 3G1 Canada University of Waterloo Department of Physics Department of Electrical and Computer Engineering Department of Systems Design Engineering 200 University Avenue West WaterlooONN2l 3G1 Canada

ISBN: (纸本)9781510645264

We use the IBMQ machine to implement a quantum database circuit. The database is made of a single multi-column table with multiple superposed records using Toffoli, CNOT and single quantum qubit gates. We present an optimized quantum circuit which allows us to select an item from the table, according to a specific oracle. Our circuit is based only on a single auxiliary qubit to perform the Selective Inversion (SI) and the Inversion Around the Average (IAA), which constitute the Grover operator. We analyze the difference between the generated transpired circuits while executing the original circuit on different IBMQ machines, namely the Melbourne and Athens quantum machines. We discuss the physical constraints causing the difference between the simulation on the composer and the executions of the circuits on the real IBMQ devices. © 2021 Copyright SPIE.

关键词： Timing circuits

When Quantum Information Technologies Meet Blockchain in Web 3.0

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Xu, Minrui Ren, Xiaoxu Niyato, Dusit Kang, Jiawen Qiu, Chao Xiong, Zehui Wang, Xiaofei Leung, Victor C.M. School of Computer Science and Engineering Nanyang Technological University Singapore College of Intelligence and Computing Tianjin University Tianjin300072 China Shenzhen518000 China School of Automation Guangdong University of Technology China Pillar of Information Systems Technology and Design Singapore University of Technology and Design Singapore487372 Singapore College of Computer Science and Software Engineering Shenzhen University Shenzhen518061 China Department of Electrical and Computer Engineering The University of British Columbia VancouverBCV6T 1Z4 Canada

With the drive to create a decentralized digital economy, Web 3.0 has become a cornerstone of digital transformation, developed on the basis of computing-force networking, distributed data storage, and blockchain. With the rapid realization of quantum devices, Web 3.0 is being developed in parallel with the deployment of quantum cloud computing and quantum Internet. In this regard, quantum computing first disrupts the original cryptographic systems that protect data security while reshaping modern cryptography with the advantages of quantum computing and communication. Therefore, in this paper, we introduce a quantum blockchain-driven Web 3.0 framework that provides information-theoretic security for decentralized data transferring and payment transactions. First, we present the framework of quantum blockchain-driven Web 3.0 with future-proof security during the transmission of data and transaction information. Next, we discuss the potential applications and challenges of implementing quantum blockchain in Web 3.0. Finally, we describe a use case for quantum non-fungible tokens (NFTs) and propose a quantum deep learning-based optimal auction for NFT trading to maximize the achievable revenue for sufficient liquidity in Web 3.0. In this way, the proposed framework can achieve proven security and sustainability for the next-generation decentralized digital society. © 2022, CC BY.

关键词： Blockchain

A Full Dive into Realizing the Edge-enabled Metaverse: Visions, Enabling Technologies,and Challenges

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Xu, Minrui Ng, Wei Chong Lim, Wei Yang Bryan Kang, Jiawen Xiong, Zehui Niyato, Dusit Yang, Qiang Shen, Xuemin Sherman Miao, Chunyan The School of Computer Science and Engineering Nanyang Technological University Singapore Alibaba Group Alibaba-NTU Joint Research Institute Nanyang Technological University Singapore The School of Automation Guangdong University of Technology China The Pillar of Information Systems Technology and Design Singapore University of Technology and Design Singapore487372 Singapore The Department of Computer Science and Engineering Hong Kong University of Science and Technology Hong Kong Webank China The Department of Electrical and Computer Engineering University of Waterloo WaterlooONN2L 3G1 Canada

Dubbed "the successor to the mobile Internet", the concept of the Metaverse has grown in popularity. While there exist lite versions of the Metaverse today, they are still far from realizing the full vision of an immersive, embodied, and interoperable Metaverse. Without addressing the issues of implementation from the communication and networking, as well as computation perspectives, the Metaverse is difficult to succeed the Internet, especially in terms of its accessibility to billions of users today. In this survey, we focus on the edge-enabled Metaverse to realize its ultimate vision. We first provide readers with a succinct tutorial of the Metaverse, an introduction to the architecture, as well as current developments. To enable ubiquitous, seamless, and embodied access to the Metaverse, we discuss the communication and networking challenges and survey cutting-edge solutions and concepts that leverage next-generation communication systems for users to immerse as and interact with embodied avatars in the Metaverse. Moreover, given the high computation costs required, e.g., to render 3D virtual worlds and run data-hungry artificial intelligence-driven avatars, we discuss the computation challenges and cloud-edge-end computation framework-driven solutions to realize the Metaverse on resource-constrained edge devices. Next, we explore how blockchain technologies can aid in the interoperable development of the Metaverse, not just in terms of empowering the economic circulation of virtual user-generated content but also to manage physical edge resources in a decentralized, transparent, and immutable manner. Finally, we discuss the future research directions towards realizing the true vision of the edge-enabled Metaverse. © 2022, CC BY.

关键词： Blockchain