220 Program Search with Simulated Annealing
Yoshihisa FUJITA, Mitsunori Miki, Masafumi Hashimoto, Tomoyuki Hiroyasu · The Proceedings of OPTIS · 2006
In this research, we verify the effectiveness of Simulated Annealing(SA) in program search by paying attention to the temperature parameter which controls the transition probability of bad moves. It is found that the temperature can control the bloat which is the big problem in Genetic Programming(GP). The fixed temperature search with important temperature instead of conventional cooling schedule provided the same performance as GP without the occurring bloat.