Robust Inverse Reinforcement Learning under Transition Dynamics Mismatch

Authors: Luca Viano, Yu-Ting Huang, Parameswaran Kamalaruban, Adrian Weller, Volkan Cevher

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we empirically demonstrate the stable performance of our algorithm compared to the standard MCE IRL algorithm under transition dynamics mismatches in both finite and continuous MDP problems.
Researcher Affiliation Collaboration Luca Viano LIONS, EPFL Yu-Ting Huang EPFL Parameswaran Kamalaruban The Alan Turing Institute Adrian Weller University of Cambridge & The Alan Turing Institute Volkan Cevher LIONS, EPFL
Pseudocode Yes Algorithm 1 Robust MCE IRL via Markov Game
Open Source Code Yes Code Repository https://github.com/lviano/Robust_MCE_IRL/tree/master/robust_IRLcode
Open Datasets No The paper describes generating data within custom GRIDWORLD and OBJECTWORLD environments, rather than using a pre-existing publicly available dataset with a specific link or citation for access. It defines how the environments are set up but does not provide concrete access information for a dataset.
Dataset Splits No The paper specifies experimental parameters like noise levels (ϵL, ϵE) and algorithm parameter α, but it does not describe specific train/validation/test dataset splits with percentages, sample counts, or references to predefined splits.
Hardware Specification No We used an internal cluster with CPU nodes for the experiments; but we do not have an estimate of the total amount of compute.
Software Dependencies No The paper mentions using the 'deep MCE IRL algorithm from [44]' but does not provide specific software dependencies with version numbers (e.g., PyTorch version, Python version, specific library versions).
Experiment Setup Yes We have provided all the training and hyperparameters details in the Experiments section, and in the Appendix. In our experiments, we set T ref to be deterministic, and T to be uniform. Then, one can easily show that ddyn T L,ϵL, T E,ϵE = 2 1 1 |S| |ϵL ϵE|. ...our robust MCE IRL algorithm with different values of α {0.8, 0.85, 0.9, 0.95}...