An Efficient Ant Colony Programming Approach

In this paper, a novel ant colony optimization for linear imperative programming (ACOP) is proposed to improve the efficiency and accuracy of automating the design of computer programs. Different from existing linear genetic programming (LGP), the evolution of ACOP is based on cooperation of artificial ants. In ACOP, each solution is a sequence of instructions. The ants treat elements (registers or operators) in instructions as nodes in the construction graph. An ant chooses its next element according to the amount of pheromone deposited during the generation of a solution. The performance of ACOP is tested on twelve benchmark symbolic regression problems. Experimental results show that ACOP can perform better or competitive in comparison with two well-known genetic programming variants.