Identifiability and Generalizability from multiple experts in inverse reinforcement learning
Code for the NeurIPS 2022 paper "Identifiability and Generalizability from multiple experts in inverse reinforcement learning"
All the results in the paper can be reproduced with the notebook Experiments.ipynb
The folder src/ contains the code for the gridworld environment and the regularized MDP solver to compute entropy regularized optimal policies.
The file functions.py contains various functions that are invoked in Experiments.ipynb