Patching Approximate Solutions in Reinforcement Learning

Abstract

This paper introduces an approach to improving an approximate solution in reinforcement learning by augmenting it with a small overriding patch. Many approximate solutions are smaller and easier to produce than a flat solution, but the best solution within the constraints of the approximation may fall well short of global optimality. We present a technique for efficiently learning a small patch to reduce this gap. Empirical evaluation demonstrates the effectiveness of patching, producing combined solutions that are much closer to global optimality.

Cite

Text

Kim and Uther. "Patching Approximate Solutions in Reinforcement Learning." European Conference on Machine Learning, 2006. doi:10.1007/11871842_27

Markdown

[Kim and Uther. "Patching Approximate Solutions in Reinforcement Learning." European Conference on Machine Learning, 2006.](https://mlanthology.org/ecmlpkdd/2006/kim2006ecml-patching/) doi:10.1007/11871842_27

BibTeX

@inproceedings{kim2006ecml-patching,
  title     = {{Patching Approximate Solutions in Reinforcement Learning}},
  author    = {Kim, Min Sub and Uther, William T. B.},
  booktitle = {European Conference on Machine Learning},
  year      = {2006},
  pages     = {258-269},
  doi       = {10.1007/11871842_27},
  url       = {https://mlanthology.org/ecmlpkdd/2006/kim2006ecml-patching/}
}