Repeated game analysis on ART adaptive categorization game
Wai-keung Fung, Yunhui Liu · 2000
A new formulation of an adaptive categorization game of ART-type neural networks is described in this paper. Learning automata dynamics is introduced into the adaptive categorization game in ART-type networks so that the vigilance parameter /spl rho/ is adjusted to give variable-sized clusters. Moreover, repeated game analysis is also conducted on the learning automata-based adaptive categorization game in order to study the asymptotic behaviors of the game when data patterns are passed to the ART networks one after the other.