About: The PG library is a high-performance reinforcement learning library. The name PG refers to policy-gradient methods, but this name is largely historical. The library also impliments value-based RL [...] Changes:Initial Announcement on mloss.org.
|