Meta-model Assisted Evolutionary Optimization of Cellular Automata: An Application to the SCIARA Model

The automatic optimization of Cellular Automata (CA) models often requires a large number of time-consuming simulations before an acceptable solution can be found. As a result, CA optimization processes may involve significant computational resources. In this paper we investigate the possibility of speeding up a CA calibration through the approach of meta-model assisted search, which is widely used in many fields. The adopted technique relies on inexpensive surrogate functions able to approximate the fitness corresponding to the CA simulations. The calibration exercise presented here refers to SCIARA, a CA for the simulation of lava flows. According to the preliminary results, the use of meta-models enables to achieve a significant gain in computational time.