Federated Reinforcement Learning at the Edge

Konstantinos Gatsis

Federated Reinforcement Learning at the Edge

Machine Learning 2021-12-14 v1 Systems and Control Systems and Control

Authors: Konstantinos Gatsis

Abstract

Modern cyber-physical architectures use data collected from systems at different physical locations to learn appropriate behaviors and adapt to uncertain environments. However, an important challenge arises as communication exchanges at the edge of networked systems are costly due to limited resources. This paper considers a setup where multiple agents need to communicate efficiently in order to jointly solve a reinforcement learning problem over time-series data collected in a distributed manner. This is posed as learning an approximate value function over a communication network. An algorithm for achieving communication efficiency is proposed, supported with theoretical guarantees, practical implementations, and numerical evaluations. The approach is based on the idea of communicating only when sufficiently informative data is collected.

Keywords

multi-agent reinforcement learning reinforcement learning federated learning

Cite

@article{arxiv.2112.05908,
  title  = {Federated Reinforcement Learning at the Edge},
  author = {Konstantinos Gatsis},
  journal= {arXiv preprint arXiv:2112.05908},
  year   = {2021}
}

Comments

arXiv admin note: text overlap with arXiv:2103.04140

Federated Reinforcement Learning at the Edge

Abstract

Keywords

Cite

Comments

Related papers