Jornal de Ciência da Computação e Biologia de Sistemas

Grammatical Herding

Abstract

Chris Headleand and William J. Teahan

Automatic programming algorithms have often looked towards biology to provide inspiration for how agents can learn to solve problems. Evolutionary and swarm based methods in particular have shown great promise in how this objective can be achieved. We present Grammatical Herding, a new fitness-based automatic programming algorithm based on a simple set of rules inspired by the herd movements of horses. In this paper, we establish the design of the new algorithm and test it against a standard benchmark problem, the Santa Fe Trail.