Min Max Generalization for Two-stage Deterministic Batch Mode Reinforcement Learning: Relaxation Schemes
Systems and Control
2012-10-31 v2 Machine Learning
Abstract
We study the minmax optimization problem introduced in [22] for computing policies for batch mode reinforcement learning in a deterministic setting. First, we show that this problem is NP-hard. In the two-stage case, we provide two relaxation schemes. The first relaxation scheme works by dropping some constraints in order to obtain a problem that is solvable in polynomial time. The second relaxation scheme, based on a Lagrangian relaxation where all constraints are dualized, leads to a conic quadratic programming problem. We also theoretically prove and empirically illustrate that both relaxation schemes provide better results than those given in [22].
Cite
@article{arxiv.1202.5298,
title = {Min Max Generalization for Two-stage Deterministic Batch Mode Reinforcement Learning: Relaxation Schemes},
author = {Raphael Fonteneau and Damien Ernst and Bernard Boigelot and Quentin Louveaux},
journal= {arXiv preprint arXiv:1202.5298},
year = {2012}
}