An implementation of the reinforcement learning agent used to generate the results in A Monte-Carlo AIXI Approximation. See my publications section for more information.
A simplified implementation was also constructed by Marcus Hutter and his students (in particular Daniel Visentin) in their 2010 Advanced AI class. It can be found here.
v1.0 CTS source and Windows binaries
An implementation of Context Tree Switching (CTS), a new general purpose data compressor. See here for a technical description.
A small MCTS-based Go program, written quickly in the second half of 2007 for the purposes of gaining some practical experience with the UCT algorithm in a competitive setting. Competed in a few KGS tournaments (28th, 29th, 31st) under the handle agog, with a best place finish of 2nd. Rated at 2319±18 Elo on the CGOS 9x9 Go Server.
A generic alpha-beta search chess program I developed as an undergraduate student in 2005. Played thousands of games on the Internet Chess Club under the handle bodo. Competed in a number of online computer chess tournaments, with the best results being a 1st from 12 in the (Australian) NC3 2005, and an equal 17th/54 in the (world-wide) CCT6 and CCT8 events.
mc-aixi v2.0 underway...
CTS v1.0 released!