Robust Decentralized Learning Using ADMM With Unreliable Agents

Li, Qunwei; Kailkhura, Bhavya; Goldhahn, Ryan; Ray, Priyadip; Varshney, Pramod K.

doi:10.1109/tsp.2022.3178655

Robust Decentralized Learning Using ADMM With Unreliable Agents

Journal Article · Fri Jun 03 00:00:00 EDT 2022 · IEEE Transactions on Signal Processing

DOI:https://doi.org/10.1109/tsp.2022.3178655· OSTI ID:1959428

^[1]; ^[1]; Goldhahn, Ryan ^[1]; ^[1]; ^[2]

Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
Syracuse University, NY (United States)

Many signal processing and machine learning problems can be formulated as consensus optimization problems which can be solved efficiently via a cooperative multi-agent system. However, the agents in the system can be unreliable due to a variety of reasons: noise, faults and attacks. Providing erroneous updates leads the optimization process in a wrong direction, and degrades the performance of distributed machine learning algorithms. This paper considers the problem of decentralized learning using ADMM in the presence of unreliable agents. First, we rigorously analyze the effect of erroneous updates (in ADMM learning iterations) on the convergence behavior of the multi-agent system. We show that the algorithm linearly converges to a neighborhood of the optimal solution under certain conditions and characterize the neighborhood size analytically. Next, we provide guidelines for network design to achieve a faster convergence to the neighborhood. Here, we also provide conditions on the erroneous updates for exact convergence to the optimal solution. Finally, to mitigate the influence of unreliable agents, we propose ROAD , a robust variant of ADMM, and show its resilience to unreliable agents with an exact convergence to the optimum.

View Accepted Manuscript (DOE)

Research Organization:: Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)

Sponsoring Organization:: USDOE Laboratory Directed Research and Development (LDRD) Program; USDOE National Nuclear Security Administration (NNSA)

Grant/Contract Number:: AC52-07NA27344

OSTI ID:: 1959428

Report Number(s):: LLNL-JRNL-796199; 997251

Journal Information:: IEEE Transactions on Signal Processing, Journal Name: IEEE Transactions on Signal Processing Journal Issue: N/A Vol. 70; ISSN 1053-587X

Publisher:: IEEECopyright Statement

Country of Publication:: United States

Language:: English

References (28)

Convergence Rate of Distributed ADMM Over Networks Makhdoumi, Ali; Ozdaglar, Asuman IEEE Transactions on Automatic Control, Vol. 62, Issue 10 https://doi.org/10.1109/tac.2017.2677879	journal	October 2017
On the linear convergence of the alternating direction method of multipliers Hong, Mingyi; Luo, Zhi-Quan Mathematical Programming, Vol. 162, Issue 1-2 https://doi.org/10.1007/s10107-016-1034-2	journal	July 2016
Approximate ADMM algorithms derived from Lagrangian splitting Eckstein, Jonathan; Yao, Wang Computational Optimization and Applications, Vol. 68, Issue 2 https://doi.org/10.1007/s10589-017-9911-z	journal	April 2017
An Inexact Alternating Directions Algorithm for Constrained Total Variation Regularized Compressive Sensing Problems Xiao, Yun-Hai; Song, Hui-Na Journal of Mathematical Imaging and Vision, Vol. 44, Issue 2 https://doi.org/10.1007/s10851-011-0314-y	journal	August 2011
Inexact Alternating-Direction-Based Contraction Methods for Separable Linearly Constrained Convex Optimization Gu, Guoyong; He, Bingsheng; Yang, Junfeng Journal of Optimization Theory and Applications, Vol. 163, Issue 1 https://doi.org/10.1007/s10957-013-0489-z	journal	December 2013
An inexact alternating direction method for solving a class of structured variational inequalities Bnouhachem, Abdellah; Benazza, Hafida; Khalfaoui, Mohamed Applied Mathematics and Computation, Vol. 219, Issue 14 https://doi.org/10.1016/j.amc.2013.01.067	journal	March 2013
Decentralized transactive energy management of multi-microgrid distribution systems based on ADMM Rajaei, Ali; Fattaheian-Dehkordi, Sajjad; Fotuhi-Firuzabad, Mahmud International Journal of Electrical Power & Energy Systems, Vol. 132 https://doi.org/10.1016/j.ijepes.2021.107126	journal	November 2021
The improvement with relative errors of he et al.'s inexact alternating direction method for monotone variational inequalities Yuan, Xiao-Ming Mathematical and Computer Modelling, Vol. 42, Issue 11-12 https://doi.org/10.1016/j.mcm.2005.04.007	journal	December 2005
Differentially Private Robust ADMM for Distributed Machine Learning Ding, Jiahao; Zhang, Xinyue; Chen, Mingsong 2019 IEEE International Conference on Big Data (Big Data) https://doi.org/10.1109/BigData47090.2019.9005716	conference	December 2019
Analysis of distributed ADMM algorithm for consensus optimization in presence of error Majzoobi, Layla; Lahouti, Farshad 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) https://doi.org/10.1109/ICASSP.2016.7472595	conference	March 2016
Preconditioning ADMM for Fast Decentralized Optimization Ma, Meng; Giannakis, Georgios B. ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) https://doi.org/10.1109/ICASSP40776.2020.9054217	conference	May 2020
Privacy-Preserving Distributed Learning with Nonsmooth Objective Functions Gauthier, Francois; Gratton, Cristiano; Venkategowda, Naveen K. D. 2020 54th Asilomar Conference on Signals, Systems, and Computers https://doi.org/10.1109/IEEECONF51394.2020.9443287	conference	November 2020
Distributed Alternating Direction Multiplier Method Based on Optimized Topology and Nodes Selection Strategy Zeng, Shuai; Lin, Haitao 2020 3rd International Seminar on Research of Information Technology and Intelligent Systems (ISRITI) https://doi.org/10.1109/ISRITI51436.2020.9315416	conference	December 2020
Coded Stochastic ADMM for Decentralized Consensus Optimization With Edge Computing Chen, Hao; Ye, Yu; Xiao, Ming IEEE Internet of Things Journal, Vol. 8, Issue 7 https://doi.org/10.1109/JIOT.2021.3058116	journal	April 2021
Distributed Subgradient Methods for Multi-Agent Optimization Nedic, Angelia; Ozdaglar, Asuman IEEE Transactions on Automatic Control, Vol. 54, Issue 1 https://doi.org/10.1109/TAC.2008.2009515	journal	January 2009
Dual Averaging for Distributed Optimization: Convergence Analysis and Network Scaling Duchi, J. C.; Agarwal, A.; Wainwright, M. J. IEEE Transactions on Automatic Control, Vol. 57, Issue 3 https://doi.org/10.1109/TAC.2011.2161027	journal	March 2012
Convergence Rate of Distributed ADMM Over Networks Makhdoumi, Ali; Ozdaglar, Asuman IEEE Transactions on Automatic Control, Vol. 62, Issue 10 https://doi.org/10.1109/TAC.2017.2677879	journal	October 2017
A Collaborative Training Algorithm for Distributed Learning Predd, Joel B.; Kulkarni, Sanjeev R.; Poor, H. Vincent IEEE Transactions on Information Theory, Vol. 55, Issue 4 https://doi.org/10.1109/TIT.2009.2012992	journal	April 2009
Comparison of Decentralized ADMM Optimization Algorithms for Power Allocation in Modular Fuel Cell Vehicles Khalatbarisoltani, Arash; Kandidayeni, Mohsen; Boulon, Loic IEEE/ASME Transactions on Mechatronics, Vol. 27, Issue 5 https://doi.org/10.1109/TMECH.2021.3105950	journal	October 2022
Consensus in Ad Hoc WSNs With Noisy Links—Part II: Distributed Estimation and Smoothing of Random Signals Schizas, Ioannis D.; Giannakis, Georgios B.; Roumeliotis, Stergios I. IEEE Transactions on Signal Processing, Vol. 56, Issue 4 https://doi.org/10.1109/TSP.2007.908943	journal	April 2008
On the Linear Convergence of the ADMM in Decentralized Consensus Optimization Shi, Wei; Ling, Qing; Yuan, Kun IEEE Transactions on Signal Processing, Vol. 62, Issue 7 https://doi.org/10.1109/TSP.2014.2304432	journal	April 2014
Multi-Agent Distributed Optimization via Inexact Consensus ADMM Chang, Tsung-Hui; Hong, Mingyi; Wang, Xiangfeng IEEE Transactions on Signal Processing, Vol. 63, Issue 2 https://doi.org/10.1109/TSP.2014.2367458	journal	January 2015
DLM: Decentralized Linearized Alternating Direction Method of Multipliers Ling, Qing; Shi, Wei; Wu, Gang IEEE Transactions on Signal Processing, Vol. 63, Issue 15 https://doi.org/10.1109/TSP.2015.2436358	journal	August 2015
Inexact Alternating Direction Methods for Image Recovery Ng, Michael K.; Wang, Fan; Yuan, Xiaoming SIAM Journal on Scientific Computing, Vol. 33, Issue 4 https://doi.org/10.1137/100807697	journal	January 2011
On the $O(1/n)$ Convergence Rate of the Douglas–Rachford Alternating Direction Method He, Bingsheng; Yuan, Xiaoming SIAM Journal on Numerical Analysis, Vol. 50, Issue 2 https://doi.org/10.1137/110836936	journal	January 2012
Optimization Methods for Large-Scale Machine Learning Bottou, Léon; Curtis, Frank E.; Nocedal, Jorge SIAM Review, Vol. 60, Issue 2 https://doi.org/10.1137/16M1080173	journal	January 2018
Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers Boyd, Stephen; Parikh, Neal; Chu, Eric https://doi.org/10.1561/2200000016	book	January 2011
Effects of network topology on the performance of consensus and distributed learning of SVMs using ADMM Tavara, Shirin; Schliep, Alexander PeerJ Computer Science, Vol. 7 https://doi.org/10.7717/peerj-cs.397	journal	March 2021

Cited By (1)

Privacy-Preserving Distributed Machine Learning via Local Randomization and ADMM Perturbation Wang, Xin; Ishii, Hideaki; Du, Linkang IEEE Transactions on Signal Processing, Vol. 68 https://doi.org/10.1109/tsp.2020.3009007	journal	January 2020

Similar Records

Distributed Reinforcement Learning with ADMM-RL

Conference · Thu Aug 29 00:00:00 EDT 2019 · OSTI ID:1669404

Decentralized Failure-Tolerant Optimization of Electric Vehicle Charging

Journal Article · Thu May 13 20:00:00 EDT 2021 · IEEE Transactions on Smart Grid · OSTI ID:1798449

Related Subjects

97 MATHEMATICS AND COMPUTING
ADMM
decentralized learning
unreliable

Robust Decentralized Learning Using ADMM With Unreliable Agents

Citation Formats

References (28)

Cited By (1)

Similar Records

Related Subjects