检索结果-内蒙古大学图书馆

MDA: Availability-Aware Federated Learning Client Selection

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Abyane, Amin Eslami Drew, Steve Hemmati, Hadi The Department of Electrical and Software Engineering University of Calgary CalgaryABT2N 1N4 Canada The Department of Electrical Engineering and Computer Science York University TorontoONM3J 1P3 Canada

Recently, a new distributed learning scheme called Federated Learning (FL) has been introduced. FL is designed so that server never collects user-owned data meaning it is great at preserving privacy. FL’s process starts with the server sending a model to clients, then the clients train that model using their data and send the updated model back to the server. Afterward, the server aggregates all the updates and modifies the global model. This process is repeated until the model converges. This study focuses on an FL setting called cross-device FL, which trains based on a large number of clients. Since many devices may be unavailable in cross-device FL, and communication between the server and all clients is extremely costly, only a fraction of clients gets selected for training at each round. In vanilla FL, clients are selected randomly, which results in an acceptable accuracy but is not ideal from the overall training time perspective, since some clients are slow and can cause some training rounds to be slow. If only fast clients get selected the learning would speed up, but it will be biased toward only the fast clients’ data, and the accuracy degrades. Consequently, new client selection techniques have been proposed to improve the training time by considering individual clients’ resources and speed. This paper introduces the first availability-aware selection strategy called MDA. The results show that our approach makes learning faster than vanilla FL by up to 6.5%. Moreover, we show that resource heterogeneity-aware techniques are effective but can become even better when combined with our approach, making it faster than the state-of-the-art selectors by up to 16%. Lastly, our approach selects more unique clients for training compared to client selectors that only select fast clients, which reduces our technique’s bias. The data that support the findings of this study are available in our repository at https://***/aminesi/FedML-Extended. © 2022, CC BY.

关键词： Software reliability

Beyond Diagonal Reconfigurable Intelligent Surfaces: A Multi-Sector Mode Enabling Highly Directional Full-Space Wireless Coverage

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Hongyu Shen, Shanpu Clerckx, Bruno The Department of Electrical and Electronic Engineering Imperial College London LondonSW7 2AZ United Kingdom The Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology Clear Water Bay Kowloon Hong Kong GrazA-8010 Austria

Reconfigurable intelligent surface (RIS) has gained much traction due to its potential to manipulate the propagation environment via nearly-passive reconfigurable elements. In our previous work, we have analyzed and proposed a beyond diagonal RIS (BD-RIS) model, which is not limited to traditional diagonal phase shift matrices, to unify different RIS modes/architectures. In this paper, we create a new branch of BD-RIS supporting a multi-sector mode. A multi-sector BD-RIS is modeled as multiple antennas connected to a multi-port group-connected reconfigurable impedance network. More specifically, antennas are divided into L (L ≥ 2) sectors and arranged as a polygon prism with each sector covering 1/L space. Different from the recently introduced concept of intelligent omni-surface (or simultaneously transmitting and reflecting RIS), the multi-sector BD-RIS not only achieves a full-space coverage, but also has significant performance gains thanks to the highly directional beam of each sector. We derive the constraint of the multi-sector BD-RIS and the corresponding channel model taking into account the relationship between antenna beamwidth and gain. With the proposed model, we first derive the scaling law of the received signal power for a multi-sector BD-RIS-assisted single-user system. We then propose efficient beamforming design algorithms to maximize the sum-rate of the multi-sector BD-RIS-assisted multiuser system. Simulation results verify the effectiveness of the proposed design and demonstrate the performance enhancement of the proposed multi-sector BD-RIS. Copyright © 2022, The Authors. All rights reserved.

关键词： Omnidirectional antennas

Electrochemical studies of nickel oxide nanoparticles via solution combustion method using green and chemical fuels

学校读者我要写书评

暂无评论

Sustainable Chemistry for the Environment

Sustainable Chemistry for the Environment 2024年 5卷

作者： Ahamed, N. Nasir Pattar, Jayadev Murthy, R.N. Keshava Kumar, M.R. Anil Bhoomika, V. Raghavendra, N. Ravikumar, C.R. Department of Physics Government College for women Chintamani 563125 India Department of Science School of applied sciences REVA University Bangalore India Department of Chemistry Government College for women Chintamani 563125 India Department of Chemical and Materials Engineering Gina Cody School of Engineering and Computer Science Concordia University Montreal H3G 2W1 Canada Department of Civil Engineering East West Institute of Technology VTU Bangalore 560091 India Research Centre Department of chemistry East West Institute of Technology Bangalore India

The efficient manufacture of nickel oxide nanoparticles (NiO NPs) using a novel solution combustion method that makes use of chemical (glucose) and green (Aloe vera gel extract) fuels is reported in this work. As revealed by electron microscopy, the synthesized samples were spongy, spherical, agglomerated, and porous;in contrast, the face-centered cubic crystal structure (FCC) phase for NiO NPs was corroborated by PXRD. The energy band gaps of NiO NPs are 4.09 and 4.21 eV. Using an eco-friendly NiO electrode, cyclic voltammetry investigations demonstrated good reversibility (a reduced value of E_O-E_R). Its great conductivity was confirmed by electrochemical impedance tests, which also showed a low charge-transfer resistance. Using NiO NPs, stable electrode materials for application in supercapacitors can be readily created. In comparison to chemically synthesized NiO green generated NiO in a 3-electrode setup demonstrated a capacitance of 156 Fg⁻¹ and 183 Fg⁻¹ at a scan rate of 5 Ag⁻¹, respectively, according to the GCD investigation. It also had good cycling stability, retaining more than 85% of its original capacitance after 2000 cycles. The results suggest that green produced NiO could be a cheap and practical material for use in supercapacitor applications in the future. © 2024 The Authors

关键词： Aloe vera extract Cyclic voltammetry Electrochemical impedance spectroscopy Photocatalyst Supercapacitor

Closed-Form Global Optimization of Beyond Diagonal Reconfigurable Intelligent Surfaces

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Nerini, Matteo Shen, Shanpu Clerckx, Bruno The Department of Electrical and Electronic Engineering Imperial College London LondonSW7 2AZ United Kingdom The Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology Clear Water Bay Kowloon Hong Kong Graz A 8010 Austria

Reconfigurable intelligent surfaces (RISs) allow controlling the propagation environment in wireless networks by tuning multiple reflecting elements. RISs have been traditionally realized through single connected architectures, mathematically characterized by a diagonal scattering matrix. Recently, beyond diagonal RISs (BD-RISs) have been proposed as a novel branch of RISs whose scattering matrix is not limited to be diagonal, which creates new benefits and opportunities for RISs. Efficient BD-RIS architectures have been realized based on group and fully connected reconfigurable impedance networks. However, a closed-form solution for the global optimal scattering matrix of these architectures is not yet available. In this paper, we provide such a closed-form solution proving that the theoretical performance upper bounds can be exactly achieved for any channel realization. We first consider the received signal power maximization in single-user single-input single-output (SISO) systems aided by a BD-RIS working in reflective or transmissive mode. Then, we extend our solution to single-user multiple-input multiple-output (MIMO) and multi-user multiple-input single-output (MISO) systems. We show that our algorithm is less complex than the iterative optimization algorithms employed in the previous literature. The complexity of our algorithm grows linearly (resp. cubically) with the number of RIS elements in the case of group (resp. fully) connected architectures. © 2022, CC BY.

关键词： Global optimization

A Lightweight Approach to Real-Time Speaker Diarization: From Audio Toward Audio-Visual Data Streams

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Kynych, Frantisek Cerva, Petr Zdansky, Jindrich Svendsen, Torbjørn Salvi, Giampiero Faculty of Mechatronics Informatics and Interdisciplinary Studies Technical University of Liberec Studentska 2 Liberec461 17 Czech Republic Department of Electronic Systems Norwegian University of Science and Technology Torgarden TrondheimNO-7491 Norway KTH Royal Institute of Technology School of Electrical Engineering and Computer Science Brinellvägen 8 StockholmSE-100 44 Sweden

This manuscript deals with the task of real-time speaker diarization (SD) for stream-wise data processing. Therefore, in contrast to most of the existing papers, it considers not only the accuracy but also the computational demands of individual investigated methods. We first propose a new lightweight scheme allowing us to perform speaker diarization of streamed audio data. Our approach utilizes a modified residual network with squeeze-and-excitation blocks (SE-ResNet-34) to extract speaker embeddings in an optimized way using cached buffers. These embeddings are subsequently used for voice activity detection (VAD) and block-online k-means clustering with a look-ahead *** described scheme yields results similar to the reference offline system while operating solely on a CPU with a low real-time factor (RTF) below 0.1 and a constant latency of around 5.5 seconds. In the next part of the work, our research moves toward much more demanding and complex real-time processing of audio-visual data *** this purpose, we extend the above-mentioned scheme for audio data processing by adding an audio-video module. This module utilizes SyncNet combined with visual embeddings for identity tracking. Our resulting multi-modal SD framework then combines the outputs from audio and audio-video modules by using a new overlap-based fusion strategy. It yields diarization error rates that are competitive with the existing state-of-the-art offline audio-visual methods while allowing us to process various audio-video streams, e.g., from internet or TV broadcasts, in real-time and with the same latency as for audio data processing. © 2023, The Authors. All rights reserved.

关键词： Embeddings

Blockchain technology in the smart city: a bibliometric review

学校读者我要写书评

暂无评论

Quality and Quantity 2022年第5期56卷 2875-2906页

作者： Rejeb, Abderahman Rejeb, Karim Simske, Steven J. Keogh, John G. Doctoral School of Regional Sciences and Business Administration Széchenyi István University Gyor 9026 Hungary Higher Institute of Computer Science El Manar 2 Rue Abou Raïhan El Bayrouni Ariana 2080 Tunisia Systems Engineering Department Colorado State University Fort Collins CO United States Henley Business School University of Reading Greenlands Henley-on-Thames RG9 3AU United Kingdom

Blockchain can function as a foundational technology with numerous applications in smart cities. The objective of this paper is twofold. First, it provides a detailed overview of the extant literature on blockchain applications in smart cities;second, it reveals the trends and suggests future research directions for scholars who wish to contribute to this rapidly growing field. We conducted a bibliometric review using a keyword co-occurrence network and article co-citation analysis. The analysis includes the assessment of 148 articles published between 2016 and 2020 in 76 academic journals. The review results demonstrate that the number of articles devoted to the study of blockchain applications and smart cities has increased exponentially in recent years. More importantly, the research identifies some of the most influential studies in this area. The paper discusses trends and highlights the challenges related to the deployment of blockchain in smart cities. To the authors’ best knowledge, this represents the first study to review the literature from leading journals on blockchain applications in smart cities using bibliometric techniques. © 2021, The Author(s).

关键词： Bibliometric review Blockchain technology Co-citation analysis Smart city

Test-Time Domain Adaptation by Learning Domain-Aware Batch Normalization

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Wu, Yanan Chi, Zhixiang Wang, Yang Plataniotis, Konstantinos N. Feng, Songhe Key Laboratory of Big Data & Artificial Intelligence in Transportation Ministry of Education Beijing Jiaotong University Beijing100044 China School of Computer and Information Technology Beijing Jiaotong University Beijing100044 China The Edward S Rogers Sr. ECE Department University of Toronto TorontoM5S3G8 Canada Department of Computer Science and Software Engineering Concordia University MontrealH3G2J1 Canada

Test-time domain adaptation aims to adapt the model trained on source domains to unseen target domains using a few unlabeled images. Emerging research has shown that the label and domain information is separately embedded in the weight matrix and batch normalization (BN) layer. Previous works normally update the whole network naively without explicitly decoupling the knowledge between label and domain. As a result, it leads to knowledge interference and defective distribution adaptation. In this work, we propose to reduce such learning interference and elevate the domain knowledge learning by only manipulating the BN layer. However, the normalization step in BN is intrinsically unstable when the statistics are re-estimated from a few samples. We find that ambiguities can be greatly reduced when only updating the two affine parameters in BN while keeping the source domain statistics. To further enhance the domain knowledge extraction from unlabeled data, we construct an auxiliary branch with label-independent self-supervised learning (SSL) to provide supervision. Moreover, we propose a bi-level optimization based on meta-learning to enforce the alignment of two learning objectives of auxiliary and main branches. The goal is to use the auxiliary branch to adapt the domain and benefit main task for subsequent inference. Our method keeps the same computational cost at inference as the auxiliary branch can be thoroughly discarded after adaptation. Extensive experiments show that our method outperforms the prior works on five WILDS real-world domain shift datasets. Our method can also be integrated with methods with label-dependent optimization to further push the performance boundary. Our code is available at https://***/ynanwu/MABN. © 2023, CC BY.

关键词： Time domain analysis

Computational Kerr-Ellipsometry: Quantifying Broadband Optical Nonreciprocity of Magneto-Optic Materials

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Choudhury, Vishal Khandekar, Chinmay Boddeti, Ashwin K. Jishi, Ali Erkovan, Mustafa Sentz, Tyler Kalhor, Farid Cardoso, Susana Supradeepa, V.R. Jacob, Zubin Max Planck Institute for the Science of Light Staudtstraße 2 Erlangen91058 Germany Birck Nanotechnology Center Elmore Family School of Electrical and Computer Engineering Purdue University West LafayetteIN47907 United States Department of Fundamental Sciences and Engineering Sivas University of Science and Technology Sivas Turkey INESC Microsistemas e Nanotecnologias Rua Alves Redol 9 Lisboa1000-029 Portugal Instituto Superior Tecnico Universidade de Lisboa Av. Rovisco Pais Lisboa1000 Portugal Centre for Nano Science and Engineering Indian Institute of Science Bengaluru560012 India

Characterizing the optical response of magneto-optic and magnetic materials usually relies on semi-classical models (e.g. Lorentz oscillator model) involving few parameters or models based on a detailed quantum mechanical description of the underlying response. These models typically involve a few parameters that are estimated via fitting the experimental data to provide a qualitative understanding of the underlying physics. Such a few-parameters fitting approach falls short of accurately capturing all elements of the complex-valued permittivity tensor across a range of wavelengths. Accurate characterization of the permittivity tensor elements across a broad range of wavelengths is invariably imperative for designing optical elements such as isolators, circulators, etc. Here, we propose and demonstrate a ubiquitous and accessible method based on a combination of spectroscopic ellipsometry and spectroscopic Magneto-Optic Kerr Effect (MOKE) measurements coupled with rigorous numerical parameter extraction techniques. To this end, we use the combined MOKE ellipsometry measurements conducted at different angles of incidence with a gradient-descent minimization algorithm to provide the inverse solution to the complete dielectric permittivity tensor. Further, we demonstrate model re-verification to ensure the estimated dielectric permittivity values reliably predict the measured experimental data. Our method is a simplified bench-top counterpart to the otherwise complex measurement systems. © 2024, CC BY-NC-SA.

关键词： Spectroscopic ellipsometry

Sherlock in OSS: A Novel Approach of Content-Based Searching in Object Storage System

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Noor, Jannatun Ratul, Rizwanul Haque Uday, Mir Rownak Ali Mondal, Joyanta Jyoti Sakif, Md. Sadiqul Islam Islam, A.B.M. Alim Al The Department of CSE Bangladesh University of Engineering and Technology Dhaka Bangladesh Research Group School of Data and Sciences BRAC University Dhaka Bangladesh C2SG Research Group Optimizely Dhaka Bangladesh C2SG Research Group Bangladesh The Department of Computer Science University of Alabama Birmingham United States mPower-Social Enterprise Dhaka Bangladesh Next-generation Computing Research Group Department of CSE Bangladesh University of Engineering and Technology Dhaka Bangladesh

Object Storage Systems (OSS) inside a cloud promise scalability, durability, availability, and concurrency. However, open-source OSS does not have a specific approach to letting users and administrators search based on the data, which is contained inside the object storage, without involving the entire cloud infrastructure. Therefore, in this paper, we propose Sherlock, a novel Content-Based Searching (CoBS) architecture to extract additional information from images and documents. Here, we store the additional information in an Elasticsearch-enabled database, which helps us to search for our desired data based on its contents. This approach works in two sequential stages. First, the data will be uploaded to a classifier that will determine the data type and send it to the specific model for the data. Here, the images that are being uploaded are sent to our trained model for object detection, and the documents are sent for keyword extraction. Next, the extracted information is sent to Elasticsearch, which enables searching based on the contents. Because the precision of the models is so fundamental to the search’s correctness, we train our models with comprehensive datasets (Microsoft COCO Dataset for multimedia data and SemEval2017 Dataset for document data). Furthermore, we put our designed architecture to the test with a real-world implementation of an open-source OSS called OpenStack Swift. We upload images into the dataset of our implementation in various segments to find out the efficacy of our proposed model in real-life Swift object storage. © 2023, CC BY-NC-SA.

关键词： Object detection