Learning Action-based Representations Using Invariance

arXiv:2403.16369v1 Announce Type: new
Abstract: Robust reinforcement learning agents using high-dimensional observations must be able to identify relevant state features amidst many exogeneous distractors. A representation that captures controllability identifies these state elements by determining what affects agent control. While methods such as inverse dynamics and mutual information capture controllability for a limited number of timesteps, capturing long-horizon elements remains a challenging problem. Myopic controllability can capture the moment right before an agent crashes into a wall, but not the …

abstract agent agents arxiv control cs.ai cs.lg dynamics features identify information reinforcement reinforcement learning representation robust state stat.ml type