readme.txt

This software package consists of a simple implementation of
MC-AIXI-CTW, an intelligent agent that learns from experience
how to perform well in a wide variety of environments. This
includes, but is not limited to the example games provided in
this package, such as Tic Tac Toe, Pacman, and Kuhn Poker.

The package consists of:
* tutorial: contains a brief tutorial to using the MC-AIXI-CTW software
* src: contains the source code for the agent
* conf: contains configuration files
* doc: contains the automatically generated documentation
  doc/html/index.html: starting point for HTML documentation
* graph: contains generated graphs
* log: contains log files. The project comes with log files generated by allowing the agent to explore for 5000 cycles then evaluating the performance of the learned policy for another 5000 cycles.

The latest version of this software can be downloaded from
http://www.hutter1.net/publ/aixictwxcode.zip

A scientific motivation and description of MC-AIXI-CTW can be found in

J.Veness, K.S.Ng, M.Hutter, W.Uther, D.Silver,
A Monte Carlo AIXI Approximation,
Journal of Artificial Intelligence Research, 40 (2011) 95-142
http://dx.doi.org/10.1613/jair.3125
Free TechReport version: http://arxiv.org/abs/0909.0801
BibTeX: http://www.hutter1.net/official/bib.htm#aixictwx