Temporal Predictive Coding For Model-Based Planning In Latent Space

Tung D Nguyen, Rui Shu, Tuan Pham, Hung Bui, Stefano Ermon
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:8130-8139, 2021.

Abstract

High-dimensional observations are a major challenge in the application of model-based reinforcement learning (MBRL) to real-world environments. To handle high-dimensional sensory inputs, existing approaches use representation learning to map high-dimensional observations into a lower-dimensional latent space that is more amenable to dynamics estimation and planning. In this work, we present an information-theoretic approach that employs temporal predictive coding to encode elements in the environment that can be predicted across time. Since this approach focuses on encoding temporally-predictable information, we implicitly prioritize the encoding of task-relevant components over nuisance information within the environment that are provably task-irrelevant. By learning this representation in conjunction with a recurrent state space model, we can then perform planning in latent space. We evaluate our model on a challenging modification of standard DMControl tasks where the background is replaced with natural videos that contain complex but irrelevant information to the planning task. Our experiments show that our model is superior to existing methods in the challenging complex-background setting while remaining competitive with current state-of-the-art models in the standard setting.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-nguyen21h, title = {Temporal Predictive Coding For Model-Based Planning In Latent Space}, author = {Nguyen, Tung D and Shu, Rui and Pham, Tuan and Bui, Hung and Ermon, Stefano}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {8130--8139}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/nguyen21h/nguyen21h.pdf}, url = {https://proceedings.mlr.press/v139/nguyen21h.html}, abstract = {High-dimensional observations are a major challenge in the application of model-based reinforcement learning (MBRL) to real-world environments. To handle high-dimensional sensory inputs, existing approaches use representation learning to map high-dimensional observations into a lower-dimensional latent space that is more amenable to dynamics estimation and planning. In this work, we present an information-theoretic approach that employs temporal predictive coding to encode elements in the environment that can be predicted across time. Since this approach focuses on encoding temporally-predictable information, we implicitly prioritize the encoding of task-relevant components over nuisance information within the environment that are provably task-irrelevant. By learning this representation in conjunction with a recurrent state space model, we can then perform planning in latent space. We evaluate our model on a challenging modification of standard DMControl tasks where the background is replaced with natural videos that contain complex but irrelevant information to the planning task. Our experiments show that our model is superior to existing methods in the challenging complex-background setting while remaining competitive with current state-of-the-art models in the standard setting.} }
Endnote
%0 Conference Paper %T Temporal Predictive Coding For Model-Based Planning In Latent Space %A Tung D Nguyen %A Rui Shu %A Tuan Pham %A Hung Bui %A Stefano Ermon %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-nguyen21h %I PMLR %P 8130--8139 %U https://proceedings.mlr.press/v139/nguyen21h.html %V 139 %X High-dimensional observations are a major challenge in the application of model-based reinforcement learning (MBRL) to real-world environments. To handle high-dimensional sensory inputs, existing approaches use representation learning to map high-dimensional observations into a lower-dimensional latent space that is more amenable to dynamics estimation and planning. In this work, we present an information-theoretic approach that employs temporal predictive coding to encode elements in the environment that can be predicted across time. Since this approach focuses on encoding temporally-predictable information, we implicitly prioritize the encoding of task-relevant components over nuisance information within the environment that are provably task-irrelevant. By learning this representation in conjunction with a recurrent state space model, we can then perform planning in latent space. We evaluate our model on a challenging modification of standard DMControl tasks where the background is replaced with natural videos that contain complex but irrelevant information to the planning task. Our experiments show that our model is superior to existing methods in the challenging complex-background setting while remaining competitive with current state-of-the-art models in the standard setting.
APA
Nguyen, T.D., Shu, R., Pham, T., Bui, H. & Ermon, S.. (2021). Temporal Predictive Coding For Model-Based Planning In Latent Space. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:8130-8139 Available from https://proceedings.mlr.press/v139/nguyen21h.html.

Related Material