Oral

Hyperparameter Selection for Imitation Learning

LĂ©onard Hussenot · Marcin Andrychowicz · Damien Vincent · Robert Dadashi · Anton Raichuk · Sabela Ramos · Nikola Momchev · Sertan Girgin · Raphael Marinier · Lukasz Stafiniak · Emmanuel Orsini · Olivier Bachem · Matthieu Geist · Olivier Pietquin

[ Abstract ] [ Livestream: Visit Reinforcement Learning 9 ] [ Paper ]
Tue 20 Jul 7 p.m. — 7:20 p.m. PDT

We address the issue of tuning hyperparameters (HPs) for imitation learning algorithms in the context of continuous-control, when the underlying reward function of the demonstrating expert cannot be observed at any time. The vast literature in imitation learning mostly considers this reward function to be available for HP selection, but this is not a realistic setting. Indeed, would this reward function be available, it could then directly be used for policy training and imitation would not be necessary. To tackle this mostly ignored problem, we propose a number of possible proxies to the external reward. We evaluate them in an extensive empirical study (more than 10'000 agents across 9 environments) and make practical recommendations for selecting HPs. Our results show that while imitation learning algorithms are sensitive to HP choices, it is often possible to select good enough HPs through a proxy to the reward function.

Chat is not available.