ICML 2018oral193 citations
Lipschitz Continuity in Model-based Reinforcement Learning
Kavosh Asadi, Dipendra Misra, Michael Littman
Abstract
We examine the impact of learning Lipschitz continuous models in the context of model-based reinforcement learning. We provide a novel bound on multi-step prediction error of Lipschitz models where we quantify the error using the Wasserstein metric. We go on to prove an error bound for the value-function estimate arising from Lipschitz models and show that the estimated value function is itself Lipschitz. We conclude with empirical results that show the benefits of controlling the Lipschitz constant of neural-network models.
BibTeX
@InProceedings{pmlr-v80-asadi18a,
title = {{L}ipschitz Continuity in Model-based Reinforcement Learning},
author = {Asadi, Kavosh and Misra, Dipendra and Littman, Michael},
booktitle = {Proceedings of the 35th International Conference on Machine Learning},
pages = {264--273},
year = {2018},
editor = {Dy, Jennifer and Krause, Andreas},
volume = {80},
series = {Proceedings of Machine Learning Research},
month = {10--15 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v80/asadi18a/asadi18a.pdf},
url = {https://proceedings.mlr.press/v80/asadi18a.html},
abstract = {We examine the impact of learning Lipschitz continuous models in the context of model-based reinforcement learning. We provide a novel bound on multi-step prediction error of Lipschitz models where we quantify the error using the Wasserstein metric. We go on to prove an error bound for the value-function estimate arising from Lipschitz models and show that the estimated value function is itself Lipschitz. We conclude with empirical results that show the benefits of controlling the Lipschitz constant of neural-network models.}
}