PIQLE was primarily designed for implementing and testing the algorithms and problems described in Sutton's and Barto's Reinforcement Learning, An Introduction
Reinforcement learning algorithms are relatively generic, as based on a quite abstract notions of state, action,reward.
We tried to respect the genericity and transcript it in Java, a language very well suited for this.
Defining and designing separately the notions of agent, algorithms, environment, and how those objects communicate with each other made it possible to obtain a very general platform for reinforcement learning experiments , easy to understand, easy to use, and also easy to extend, by mean of adding new algorithms, new problems, new agents...
- Changes to previous version:
Initial Announcement on mloss.org.
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.