Bayesian Inverse Transition Learning: Learning Dynamics from Near-Optimal Trajectories

Leo Benac, Abhishek Sharma, Sonali Parbhoo, Finale Doshi-Velez
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:2242-2250, 2026.

Abstract

We consider the problem of estimating the transition dynamics from near-optimal expert trajectories in the context of offline model-based reinforcement learning. We develop a novel constraint-based method, Inverse Transition Learning, that treats the limited coverage of the expert trajectories as a feature: we use the fact that the expert is near-optimal to inform our estimate of. We integrate our constraints into a Bayesian approach. Across both synthetic environments and real healthcare scenarios like Intensive Care Unit (ICU) patient management in hypotension, we demonstrate not only significant improvements in decision-making, but that our posterior can inform when transfer will be successful.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-benac26a, title = { Bayesian Inverse Transition Learning: Learning Dynamics from Near-Optimal Trajectories }, author = {Benac, Leo and Sharma, Abhishek and Parbhoo, Sonali and Doshi-Velez, Finale}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {2242--2250}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/benac26a/benac26a.pdf}, url = {https://proceedings.mlr.press/v300/benac26a.html}, abstract = { We consider the problem of estimating the transition dynamics from near-optimal expert trajectories in the context of offline model-based reinforcement learning. We develop a novel constraint-based method, Inverse Transition Learning, that treats the limited coverage of the expert trajectories as a feature: we use the fact that the expert is near-optimal to inform our estimate of. We integrate our constraints into a Bayesian approach. Across both synthetic environments and real healthcare scenarios like Intensive Care Unit (ICU) patient management in hypotension, we demonstrate not only significant improvements in decision-making, but that our posterior can inform when transfer will be successful. } }
Endnote
%0 Conference Paper %T Bayesian Inverse Transition Learning: Learning Dynamics from Near-Optimal Trajectories %A Leo Benac %A Abhishek Sharma %A Sonali Parbhoo %A Finale Doshi-Velez %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-benac26a %I PMLR %P 2242--2250 %U https://proceedings.mlr.press/v300/benac26a.html %V 300 %X We consider the problem of estimating the transition dynamics from near-optimal expert trajectories in the context of offline model-based reinforcement learning. We develop a novel constraint-based method, Inverse Transition Learning, that treats the limited coverage of the expert trajectories as a feature: we use the fact that the expert is near-optimal to inform our estimate of. We integrate our constraints into a Bayesian approach. Across both synthetic environments and real healthcare scenarios like Intensive Care Unit (ICU) patient management in hypotension, we demonstrate not only significant improvements in decision-making, but that our posterior can inform when transfer will be successful.
APA
Benac, L., Sharma, A., Parbhoo, S. & Doshi-Velez, F.. (2026). Bayesian Inverse Transition Learning: Learning Dynamics from Near-Optimal Trajectories . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:2242-2250 Available from https://proceedings.mlr.press/v300/benac26a.html.

Related Material