Emerging collective intelligence in Othello players evolved by differential evolution

Abstract

The evaluation function for game playing is very important. However, it is difficult to make a good evaluation function. In this study, we propose to play Othello using collective intelligence of players. The evaluation functions of the players are learned or optimized by Differential Evolution. The objective value is defined based on the total score of the games with a standard Othello player. In order to generate different types of players, the objective value is slightly changed by introducing the stability of each player. Each player can select a next move using the learned evaluation function. The collective intelligence player selects a move based on majority vote where the move voted by many players is selected. It is shown that the collective intelligence is effective to game players through computer simulation.