Eylam Tagor

Scalable Causal Imitation Learning

Eylam Tagor, Mingxuan Li, Elias Bareinboim

Imitation learning enables learning a policy in an unknown environment with a latent reward signal using expert demonstrations, but it struggles when the imitator's and expert's observations are mismatched and unobserved confounders are present in expert demonstrations. By identi…Read more