Reinforcement Learning with History Dependent Dynamic Contexts

Guy Tennenholtz, Nadav Merlis, Lior Shani, Martin Mladenov, Craig Boutilier
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:34011-34053, 2023.

Abstract

We introduce Dynamic Contextual Markov Decision Processes (DCMDPs), a novel reinforcement learning framework for history-dependent environments that generalizes the contextual MDP framework to handle non-Markov environments, where contexts change over time. We consider special cases of the model, with a focus on logistic DCMDPs, which break the exponential dependence on history length by leveraging aggregation functions to determine context transitions. This special structure allows us to derive an upper-confidence-bound style algorithm for which we establish regret bounds. Motivated by our theoretical results, we introduce a practical model-based algorithm for logistic DCMDPs that plans in a latent space and uses optimism over history-dependent features. We demonstrate the efficacy of our approach on a recommendation task (using MovieLens data) where user behavior dynamics evolve in response to recommendations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-tennenholtz23a, title = {Reinforcement Learning with History Dependent Dynamic Contexts}, author = {Tennenholtz, Guy and Merlis, Nadav and Shani, Lior and Mladenov, Martin and Boutilier, Craig}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {34011--34053}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/tennenholtz23a/tennenholtz23a.pdf}, url = {https://proceedings.mlr.press/v202/tennenholtz23a.html}, abstract = {We introduce Dynamic Contextual Markov Decision Processes (DCMDPs), a novel reinforcement learning framework for history-dependent environments that generalizes the contextual MDP framework to handle non-Markov environments, where contexts change over time. We consider special cases of the model, with a focus on logistic DCMDPs, which break the exponential dependence on history length by leveraging aggregation functions to determine context transitions. This special structure allows us to derive an upper-confidence-bound style algorithm for which we establish regret bounds. Motivated by our theoretical results, we introduce a practical model-based algorithm for logistic DCMDPs that plans in a latent space and uses optimism over history-dependent features. We demonstrate the efficacy of our approach on a recommendation task (using MovieLens data) where user behavior dynamics evolve in response to recommendations.} }
Endnote
%0 Conference Paper %T Reinforcement Learning with History Dependent Dynamic Contexts %A Guy Tennenholtz %A Nadav Merlis %A Lior Shani %A Martin Mladenov %A Craig Boutilier %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-tennenholtz23a %I PMLR %P 34011--34053 %U https://proceedings.mlr.press/v202/tennenholtz23a.html %V 202 %X We introduce Dynamic Contextual Markov Decision Processes (DCMDPs), a novel reinforcement learning framework for history-dependent environments that generalizes the contextual MDP framework to handle non-Markov environments, where contexts change over time. We consider special cases of the model, with a focus on logistic DCMDPs, which break the exponential dependence on history length by leveraging aggregation functions to determine context transitions. This special structure allows us to derive an upper-confidence-bound style algorithm for which we establish regret bounds. Motivated by our theoretical results, we introduce a practical model-based algorithm for logistic DCMDPs that plans in a latent space and uses optimism over history-dependent features. We demonstrate the efficacy of our approach on a recommendation task (using MovieLens data) where user behavior dynamics evolve in response to recommendations.
APA
Tennenholtz, G., Merlis, N., Shani, L., Mladenov, M. & Boutilier, C.. (2023). Reinforcement Learning with History Dependent Dynamic Contexts. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:34011-34053 Available from https://proceedings.mlr.press/v202/tennenholtz23a.html.

Related Material