Good news from Budapest, Hungary where researchers at Eötvös University have taught a computer to play the popular video game Ms Pac-Man, and quire well.

The authors developed a number of algorithms that try to learn how best to beat Pac-Man, one of which, called CE-fixedRB, outplayed the average human (but not the best human). Not surprisingly, CE-fixedRB came to the realization that it’s best to avoid ghosts (but only if they’re close-by) otherwise to continue eating dots, esp blue ones.

Their conclusion: they like the Cross-Entropy Method for simulation and optimization.

Journal of Artificial Intelligence Research - Vol 30, 12/07 - via newscientist

This entry was posted on Tuesday, January 22nd, 2008 at 11:42 pm and is filed under General. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.