Privacy-Preserving Distributed ADMM With Event-Triggered Communication

Zhang, Zhen; Yang, Shaofu; Xu, Wenying; Di, Kai

doi:10.1109/tnnls.2022.3192346

Cited by 8 publications

(4 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In DQM, information is required to be transmitted at every iteration, which is undesirable for settings where the communication source is limited. To relieve the communication burden of DQM, in our prior work [23], a communication-censored mechanism is leveraged to reduce the communication round. In order to further reduce communication costs, we not only schedule the communication instants by communication-censored mechanism but also compress the exchanged information.…”

Section: Algorithm Developmentmentioning

confidence: 99%

“…Compared with the quantized ADMM [13,24], CC-DQM can converge exactly and enjoy a smaller computation cost since it does not need to solve a subproblem at every iteration. Compared with the communication-censored ADMM [21,23,28], CC-DQM can reduce the transmitted bits per communication, thus relieving the communication cost greatly. Moreover, as we will show later, compared with the quantized first-order method [17,25], CC-DQM enjoys a faster convergence rate, thus achieving a smaller communication cost.…”

Section: Algorithm Developmentmentioning

confidence: 99%

“…• Based on ADMM, We develop a communicationefficient second-order algorithm by combining communication compression with event-triggered communication termed CC-DQM. Compared with our prior work C-DQM [23], an event-triggered communication algorithm, CC-DQM can save the transmitted bits per communication round. Compared with the existing quantized ADMM [13,24], CC-DQM can achieve an exact convergence due to the implementation of a totally different contractive compressor.…”

Section: Introductionmentioning

confidence: 98%

“…COCA [21] is a communication-censored ADMM, where agents need to solve subproblems at every iteration. C-DQM [23] is the communication-censored version of DQM. In this experiment, the compressor we implement is Example 1, the deterministic quantizer.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Decentralized ADMM with Compressed and Event-Triggered Communication

Zhang¹,

Yang²,

Xu³

2023

Preprint

View full text Add to dashboard Cite

This paper focuses on the decentralized optimization problem, where agents in a network cooperate to minimize the sum of their local objective functions by information exchange and local computation. Based on alternating direction method of multipliers (ADMM), we propose CC-DQM, a communication-efficient decentralized second-order optimization algorithm that combines compressed communication with event-triggered communication. Specifically, agents are allowed to transmit the compressed message only when the current primal variables have changed greatly compared to its last estimate. Moreover, to relieve the computation cost, the update of Hessian is scheduled by the trigger condition. To maintain exact linear convergence under compression, we compress the difference between the information to be transmitted and its estimate by a general contractive compressor. Theoretical analysis shows that CC-DQM can still achieve an exact linear convergence, despite the existence of compression error and intermittent communication, if the objective functions are strongly convex and smooth. Finally, we validate the performance of CC-DQM by numerical experiments.

show abstract

Section: Algorithm Developmentmentioning

confidence: 99%