Article
Computer Science, Information Systems
Yaser Mansouri, Victor Prokhorenko, Faheem Ullah, Muhammad Ali Babar
Summary: This study experiments on various physical and virtualized computing nodes to reveal which database under which offloading scenario is more efficient in terms of energy, bandwidth, and storage consumption in edge-cloud environments.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Engineering, Electrical & Electronic
Matthias Frey, Igor Bjelakovic, Slawomir Stanczak
Summary: OTA computation involves computing distributed data functions without transmitting all data to a central point, reducing communication cost by achieving better-than-linear scaling as the number of transmitters grows. This study proposes an analog OTA computation scheme for various functions, proving error bounds for fast-fading channels and sub-Gaussian distributions. The analysis shows potential for reducing communication cost in applications like ML-based distributed anomaly detection in large wireless sensor networks.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2021)
Article
Computer Science, Information Systems
Wooseung Nam, Joohyun Lee, Ness B. Shroff, Kyunghan Lee
Summary: SyncCoding is an inter-data encoding technique that replaces bit sequences of data to be transmitted with pointers to matching sequences in shared data, resulting in improved efficiency, reduced data traffic, faster transmission, and energy savings. Evaluations show that SyncCoding outperforms existing popular encoding techniques and achieves significant gains in data size reduction after compression in both cloud storage and web browsing scenarios.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2021)
Article
Computer Science, Information Systems
Tran Phuong Thao, Mohammad Shahriar Rahman, Md Zakirul Alam Bhuiyan, Ayumu Kubota, Shinsaku Kiyomoto, Kazumasa Omote
Summary: Secret sharing schemes are commonly used in distributed storage for Big Data to protect outsourced data and ensure key management security. Existing schemes are inefficient and lack robustness, but a new scheme proposed in this paper based on Slepian-Wolf coding achieves optimal share size and enhances exact-share repair. Experiments show that the new scheme significantly reduces communication and storage costs while supporting direct share repair with lightweight XOR operations.
IEEE TRANSACTIONS ON BIG DATA
(2021)
Article
Computer Science, Hardware & Architecture
Bahman Abolhassani, John Tadrous, Atilla Eryilmaz
Summary: The study focuses on freshness-driven caching algorithm for dynamic content and provides new models and analyses. By comparing freshness-driven caching strategies with benchmark caching strategies, the research demonstrates the advantages of freshness-driven caching strategies.
IEEE-ACM TRANSACTIONS ON NETWORKING
(2022)
Article
Computer Science, Theory & Methods
Enda Yu, Dezun Dong, Xiangke Liao
Summary: This paper proposes a standard for systematically classifying communication optimization algorithms in distributed deep learning systems based on mathematical modeling, which is a novel contribution in the field. The authors categorize existing works into four categories based on communication optimization strategies and discuss potential future challenges and research directions.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Automation & Control Systems
Kan Xie, Qianqian Cai, Zhaorong Zhang, Minyue Fu
Summary: This paper introduces fast convergent distributed algorithms for weighted average consensus, offering different solutions for acyclic and cyclic graphs with low complexities and robustness to transmission adversaries. The algorithms are conceptually different from the popular graph Laplacian approach and converge much faster than it.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2021)
Article
Computer Science, Information Systems
Song Deng, Yujia Zhai, Di Wu, Dong Yue, Xiong Fu, Yi He
Summary: This paper proposes a Lightweight Dynamic Storage Algorithm based on Adaptive Encoding (LDSA-AE) for data storage in the Energy Internet (EI). The algorithm classifies data into active and inactive categories and employs different encoding methods for each category to achieve low overhead, low latency, high reliability, and high throughput. The viability and superiority of the LDSA-AE approach are substantiated through theoretical analysis and empirical study.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Giovanni Acampora, Ferdinando Di Martino, Alfredo Massa, Roberto Schiattarella, Autilia Vitiello
Summary: This paper introduces the concept of Distributed Noisy-Intermediate Scale Quantum (D-NISQ) as a reference computational model to design innovative frameworks for quantum devices to interact and solve complex problems collaboratively. Through two case studies, a multi-threaded implementation of the D-NISQ model demonstrates greater reliability in solving problems through quantum computation.
INFORMATION FUSION
(2023)
Article
Computer Science, Information Systems
Ana Moreton-Fernandez, Yuri Torres De La Sierra, Arturo Gonzalez-Escribano, Diego R. Llanos
Summary: This paper presents a method based on four combinable operators to redistribute partial domains selected by the programmer at runtime in distributed-memory systems, aiming to improve the performance of parallel algorithms operating on changing or partial domains. The experimental results show that this approach automatically generates a good load balance for generic data-distribution policies, without introducing significant performance overheads compared to tailored data redistributions directly programmed using MPI.
Article
Engineering, Electrical & Electronic
Berkay Turan, Cesar A. Uribe, Hoi-To Wai, Mahnoosh Alizadeh
Summary: In this paper, a first-order distributed optimization algorithm that is robust to Byzantine failures is proposed. The algorithm models each agent's state as a two-state Markov chain and employs three layers of defense to achieve temporal and spatial robust aggregation, as well as gradient normalization. The method is applicable for convex and smooth non-convex cost functions and provides convergence guarantees.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2022)
Article
Engineering, Electrical & Electronic
Ziwei Xuan, Krishna Narayanan
Summary: In this paper, low-delay joint source-channel coding (JSCC) schemes for transmitting discrete-time analog sources over noisy channels based on deep neural networks are designed. The design problem is treated as the optimization of an autoencoder model, and multiple scenarios are discussed. For point-to-point communication of independent and identically distributed (i.i.d) Gaussian sources and Gauss-Markov sources over additive-white Gaussian noise (AWGN) channels, the encoder and decoder are constructed using recurrent neural networks (RNNs). With minimum prior knowledge, the performance of these RNNs-based models is optimized through fine tuning techniques during training. Sinusoidal representation networks (SIRENs)-based models are proposed and optimized for three JSCC problems, including transmitting multivariate Gaussian sources over AWGN channels, transmitting i.i.d Gaussian sources with side information at the decoder, and communicating correlated sources over orthogonal Gaussian channels. The results show that these deep learning-based JSCC schemes perform comparably or better than state-of-the-art traditional schemes. The proposed scheme can be flexibly extended to different pairs of source and channel dimensions, and the encoder mappings learned exhibit interpretable structured patterns.
IEEE TRANSACTIONS ON COMMUNICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Paulo Mendes
Summary: This article analyzes suitable networking design choices to support the Metaverse and proposes a new service-centric networking approach capable of incorporating low-latency data fetching, distributed computing, and fusion of heterogeneous data types over the Cloud-to-Thing continuum.
IEEE COMMUNICATIONS MAGAZINE
(2023)
Article
Automation & Control Systems
M. Elsisi, M. Soliman
Summary: The study presents an optimized robust non-fragile PID controller using the future search algorithm (FSA) to address uncertainties in the plant model parameters and perturbations in controller gains, ensuring robust stability and non-fragility simultaneously.
Article
Computer Science, Artificial Intelligence
Daniel Auge, Julian Hille, Etienne Mueller, Alois Knoll
Summary: Biologically inspired spiking neural networks are popular in artificial intelligence due to their efficiency in solving complex problems, but industrial applications are hindered by the challenge of encoding incoming data. This paper summarizes signal encoding schemes and proposes a uniform nomenclature to clarify definitions, surveying theoretical foundations and applications.
NEURAL PROCESSING LETTERS
(2021)