Code for abstracting, evaluating, and visualizing Markov Decision Processes.
See state_abstraction/Workspace and\ Code/stateAbstraction/src/experiments/ to run experiments.
Part of our 2016 ICML paper: "Near Optimal Behavior via Approximate State Abstraction". To reproduce results, run epsilonExperiments.java (in the above experiments directory).
Authors: David Abel, D Ellis Hershkowitz, Michael L. Littman.