Description:

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 algorithms, natural actor critic, least squares policy iteration and others. It has been designed with large distributed RL systems in mind. It's also pretty fast and modular.

API documentation and examples are provided. There is a C++ template which should make it easy to implement your problem within the LibPG framework, without needing to know anything about RL.

What libpg does NOT provide is model based planning algorithms such as value iteration, or real-time dynamic programming, or exact policy gradient. There is limited support for belief state tracking in the simulators/Cassandra/ directory (named because we use the POMDP file format created by Anthony Cassandra).

Changes to previous version:

Initial Announcement on mloss.org.

BibTeX Entry: Download

Corresponding Paper BibTeX Entry: Download

Supported Operating Systems: Agnostic

Data Formats: None

Tags: Actor Critic, Control, Least Squares Policy Iteration, Policy Gradient, Reinforcement Learning

Archive: download here

Comments

No one has posted any comments yet. Perhaps you'd like to be the first?

Leave a comment

You must be logged in to post comments.