Learning Linear-Quadratic Regulators Efficiently with only T\sqrt{T} Regret
Alon Cohen Thanks: Technion—Israel Inst.˜of Technology and Google Tel Aviv; Tomer Koren Thanks: Google Brain, Mountain View; Yishay Mansour Thanks: Tel-Aviv University and Google Tel Aviv;
Abstract
We present the first computationally-efficient algorithm with $\smash{\widetilde{O}}(\sqrt{T})$ regret for learning in Linear Quadratic Control systems with unknown dynamics. By that, we resolve an open question of Abbasi-Yadkori and Szepesvári 2011 and Dean et al. 2018.
原文 arXiv:1902.06223;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1902.06223v2