Sequential Transfer in Reinforcement Learning with a Generative Model

Andrea Tirinzoni, Riccardo Poiani, Marcello Restelli
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:9481-9492, 2020.

Abstract

We are interested in how to design reinforcement learning agents that provably reduce the sample complexity for learning new tasks by transferring knowledge from previously-solved ones. The availability of solutions to related problems poses a fundamental trade-off: whether to seek policies that are expected to immediately achieve high (yet sub-optimal) performance in the new task or whether to seek information to quickly identify an optimal solution, potentially at the cost of poor initial behaviour. In this work, we focus on the second objective when the agent has access to a generative model of state-action pairs. First, given a set of solved tasks containing an approximation of the target one, we design an algorithm that quickly identifies an accurate solution by seeking the state-action pairs that are most informative for this purpose. We derive PAC bounds on its sample complexity which clearly demonstrate the benefits of using this kind of prior knowledge. Then, we show how to learn these approximate tasks sequentially by reducing our transfer setting to a hidden Markov model and employing spectral methods to recover its parameters. Finally, we empirically verify our theoretical findings in simple simulated domains.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-tirinzoni20a, title = {Sequential Transfer in Reinforcement Learning with a Generative Model}, author = {Tirinzoni, Andrea and Poiani, Riccardo and Restelli, Marcello}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {9481--9492}, year = {2020}, editor = {Hal Daumé III and Aarti Singh}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/tirinzoni20a/tirinzoni20a.pdf}, url = { http://proceedings.mlr.press/v119/tirinzoni20a.html }, abstract = {We are interested in how to design reinforcement learning agents that provably reduce the sample complexity for learning new tasks by transferring knowledge from previously-solved ones. The availability of solutions to related problems poses a fundamental trade-off: whether to seek policies that are expected to immediately achieve high (yet sub-optimal) performance in the new task or whether to seek information to quickly identify an optimal solution, potentially at the cost of poor initial behaviour. In this work, we focus on the second objective when the agent has access to a generative model of state-action pairs. First, given a set of solved tasks containing an approximation of the target one, we design an algorithm that quickly identifies an accurate solution by seeking the state-action pairs that are most informative for this purpose. We derive PAC bounds on its sample complexity which clearly demonstrate the benefits of using this kind of prior knowledge. Then, we show how to learn these approximate tasks sequentially by reducing our transfer setting to a hidden Markov model and employing spectral methods to recover its parameters. Finally, we empirically verify our theoretical findings in simple simulated domains.} }
Endnote
%0 Conference Paper %T Sequential Transfer in Reinforcement Learning with a Generative Model %A Andrea Tirinzoni %A Riccardo Poiani %A Marcello Restelli %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-tirinzoni20a %I PMLR %P 9481--9492 %U http://proceedings.mlr.press/v119/tirinzoni20a.html %V 119 %X We are interested in how to design reinforcement learning agents that provably reduce the sample complexity for learning new tasks by transferring knowledge from previously-solved ones. The availability of solutions to related problems poses a fundamental trade-off: whether to seek policies that are expected to immediately achieve high (yet sub-optimal) performance in the new task or whether to seek information to quickly identify an optimal solution, potentially at the cost of poor initial behaviour. In this work, we focus on the second objective when the agent has access to a generative model of state-action pairs. First, given a set of solved tasks containing an approximation of the target one, we design an algorithm that quickly identifies an accurate solution by seeking the state-action pairs that are most informative for this purpose. We derive PAC bounds on its sample complexity which clearly demonstrate the benefits of using this kind of prior knowledge. Then, we show how to learn these approximate tasks sequentially by reducing our transfer setting to a hidden Markov model and employing spectral methods to recover its parameters. Finally, we empirically verify our theoretical findings in simple simulated domains.
APA
Tirinzoni, A., Poiani, R. & Restelli, M.. (2020). Sequential Transfer in Reinforcement Learning with a Generative Model. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:9481-9492 Available from http://proceedings.mlr.press/v119/tirinzoni20a.html .

Related Material