English

Risk-Averse Learning by Temporal Difference Methods

Optimization and Control 2020-03-03 v1 Machine Learning

Abstract

We consider reinforcement learning with performance evaluated by a dynamic risk measure. We construct a projected risk-averse dynamic programming equation and study its properties. Then we propose risk-averse counterparts of the methods of temporal differences and we prove their convergence with probability one. We also perform an empirical study on a complex transportation problem.

Keywords

Cite

@article{arxiv.2003.00780,
  title  = {Risk-Averse Learning by Temporal Difference Methods},
  author = {Umit Kose and Andrzej Ruszczynski},
  journal= {arXiv preprint arXiv:2003.00780},
  year   = {2020}
}