Code used in making the undergraduate thesis "Solving Montezuma's Revenge with Planning and Reinforcement Learning".
examine_game: scripts for examining the memory and testing the options.
IW: the implementation of IW(3) on position, based on Lipovetzky, Ramirez and Geffner's.
hexq: the implementation of Sarsa for the learning results.
ALE-montezuma-modified: slightly modified ALE, notably with added rewards to MR and some methods and attributes made public.
report-TFG: the thesis report, in LaTeX.
Almost all the code here uses the Arcade Learning
is GPLv2. For simplicity, then, all the code in this repository or its
sub-repositories is GPLv2 (see
The data and document are released under a Creative Commons Attribution 4.0 International License.