| Abstract [eng] |
Games with controlled environment, fixed rules and easy to compute performance, like chess or checkers, allow ideal conditions for testing Machine learning algorithms and techniques. As most games allow multiple players, competing computer versus human player makes it easy to assess the effectiveness of the algorithms. Computers have long equaled or surpassed world class human players in various games ranging from chess to backgammon, and recently it has surpassed Go, a game thought to be one of most complex, world champion. This work analyses application of heuristic methods to Mahjong game. It is accomplished using heuristic methods, artificial neural networks, constraint satisfaction problem solving methods. The effectiveness of these methods is compared and conclusions are drawn. |
