Information-Theoretic State Space Model for Multi-View Reinforcement Learning

Hyeongjoo Hwang, Seokin Seo, Youngsoo Jang, Sungyoon Kim, Geon-Hyeong Kim, Seunghoon Hong, Kee-Eung Kim
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:14249-14282, 2023.

Abstract

Multi-View Reinforcement Learning (MVRL) seeks to find an optimal control for an agent given multi-view observations from various sources. Despite recent advances in multi-view learning that aim to extract the latent representation from multi-view data, it is not straightforward to apply them to control tasks, especially when the observations are temporally dependent on one another. The problem can be even more challenging if the observations are intermittently missing for a subset of views. In this paper, we introduce Fuse2Control (F2C), an information-theoretic approach to capturing the underlying state space model from the sequences of multi-view observations. We conduct an extensive set of experiments in various control tasks showing that our method is highly effective in aggregating task-relevant information across many views, that scales linearly with the number of views while retaining robustness to arbitrary missing view scenarios.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-hwang23c, title = {Information-Theoretic State Space Model for Multi-View Reinforcement Learning}, author = {Hwang, Hyeongjoo and Seo, Seokin and Jang, Youngsoo and Kim, Sungyoon and Kim, Geon-Hyeong and Hong, Seunghoon and Kim, Kee-Eung}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {14249--14282}, 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/hwang23c/hwang23c.pdf}, url = {https://proceedings.mlr.press/v202/hwang23c.html}, abstract = {Multi-View Reinforcement Learning (MVRL) seeks to find an optimal control for an agent given multi-view observations from various sources. Despite recent advances in multi-view learning that aim to extract the latent representation from multi-view data, it is not straightforward to apply them to control tasks, especially when the observations are temporally dependent on one another. The problem can be even more challenging if the observations are intermittently missing for a subset of views. In this paper, we introduce Fuse2Control (F2C), an information-theoretic approach to capturing the underlying state space model from the sequences of multi-view observations. We conduct an extensive set of experiments in various control tasks showing that our method is highly effective in aggregating task-relevant information across many views, that scales linearly with the number of views while retaining robustness to arbitrary missing view scenarios.} }
Endnote
%0 Conference Paper %T Information-Theoretic State Space Model for Multi-View Reinforcement Learning %A Hyeongjoo Hwang %A Seokin Seo %A Youngsoo Jang %A Sungyoon Kim %A Geon-Hyeong Kim %A Seunghoon Hong %A Kee-Eung Kim %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-hwang23c %I PMLR %P 14249--14282 %U https://proceedings.mlr.press/v202/hwang23c.html %V 202 %X Multi-View Reinforcement Learning (MVRL) seeks to find an optimal control for an agent given multi-view observations from various sources. Despite recent advances in multi-view learning that aim to extract the latent representation from multi-view data, it is not straightforward to apply them to control tasks, especially when the observations are temporally dependent on one another. The problem can be even more challenging if the observations are intermittently missing for a subset of views. In this paper, we introduce Fuse2Control (F2C), an information-theoretic approach to capturing the underlying state space model from the sequences of multi-view observations. We conduct an extensive set of experiments in various control tasks showing that our method is highly effective in aggregating task-relevant information across many views, that scales linearly with the number of views while retaining robustness to arbitrary missing view scenarios.
APA
Hwang, H., Seo, S., Jang, Y., Kim, S., Kim, G., Hong, S. & Kim, K.. (2023). Information-Theoretic State Space Model for Multi-View Reinforcement Learning. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:14249-14282 Available from https://proceedings.mlr.press/v202/hwang23c.html.

Related Material