DeepMDP: Learning Continuous Latent Space Models for Representation Learning
Carles Gelada, Saurabh Kumar, Jacob Buckman, Ofir Nachum, Marc G. Bellemare
Abstract
Many reinforcement learning (RL) tasks provide the agent with high-dimensional observations that can be simplified into low-dimensional continuous states. To formalize this process, we introduce the concept of a \texit{DeepMDP}, a parameterized latent space model that is trained via the minimization of two tractable latent space losses: prediction of rewards and prediction of the distribution over next latent states. We show that the optimization of these objectives guarantees (1) the quality of the embedding function as a representation of the state space and (2) the quality of the DeepMDP as a model of the environment. Our theoretical findings are substantiated by the experimental result that a trained DeepMDP recovers the latent structure underlying high-dimensional observations on a synthetic environment. Finally, we show that learning a DeepMDP as an auxiliary task in the Atari 2600 domain leads to large performance improvements over model-free RL.
BibTeX
@InProceedings{pmlr-v97-gelada19a,
title = {{D}eep{MDP}: Learning Continuous Latent Space Models for Representation Learning},
author = {Gelada, Carles and Kumar, Saurabh and Buckman, Jacob and Nachum, Ofir and Bellemare, Marc G.},
booktitle = {Proceedings of the 36th International Conference on Machine Learning},
pages = {2170--2179},
year = {2019},
editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan},
volume = {97},
series = {Proceedings of Machine Learning Research},
month = {09--15 Jun},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v97/gelada19a/gelada19a.pdf},
url = {https://proceedings.mlr.press/v97/gelada19a.html},
abstract = {Many reinforcement learning (RL) tasks provide the agent with high-dimensional observations that can be simplified into low-dimensional continuous states. To formalize this process, we introduce the concept of a \texit{DeepMDP}, a parameterized latent space model that is trained via the minimization of two tractable latent space losses: prediction of rewards and prediction of the distribution over next latent states. We show that the optimization of these objectives guarantees (1) the quality of the embedding function as a representation of the state space and (2) the quality of the DeepMDP as a model of the environment. Our theoretical findings are substantiated by the experimental result that a trained DeepMDP recovers the latent structure underlying high-dimensional observations on a synthetic environment. Finally, we show that learning a DeepMDP as an auxiliary task in the Atari 2600 domain leads to large performance improvements over model-free RL.}
}