Uglaševanje šahovske ocenitvene funkcije s pomočjo algoritma diferencialne evolucije : doktorska disertacija

slovenian

doktorska dela

2010

Univerza v Mariboru, Fakulteta za elektrotehniko računalništvo in informatiko

We present a Differential Evolution (DE) based approach for chess evaluation function tuning. DE with opposition-based optimization is employed and upgraded with a history mechanism. The opposition-based optimization improves efficiency of DE because of noise in the evaluation of individuals. Additionally introduced history mechanism reduces noise, because potentially good individuals played more games through several generations. New history mechanism uses an auxiliary population containing potentially good individualsfrom history of the evolutionary process. This mechanism ensures that potentially good individuals remain within the evolutionary process, eventhough they died several generations back and later can be injected back into the evolutionary process. In such manner, developed algorithm reduces noise in the evaluation of individual, reduces the possibility of overfitting and consequently improves efficiency of whole tuning process.

07.08.2023