Promoting Machine-Code Program Evolution in Asynchronous Genetic Programming

摘要

This paper focuses on an asynchronous program evolution in evolutionary computation, which is hard to evolve programs effectively unlike a synchronous program evolution that evolves individuals effectively by selecting good parents after evaluations of all individuals in each generation. To tackle this problem, we explore the mechanism that can promote an asynchronous program evolution by selecting a good individual without waiting for evaluations of all individuals. For this purpose, this paper investigates the effectiveness of the proposed mechanisms in genetic programing (GP) domain by evaluating it in the two types of problems, the arithmetic and the Boolean problems. Through the intensive experiments of the eight kinds of testbeds under the two types of problems, the following implications have been revealed: (1) the program asynchronously evolved with the proposed mechanism can be completed with mostly the same or shorter execution steps than the program asynchronously evolved without the proposed mechanism, in particular the proposed mechanism improves the performance of the asynchronous evolution in the arithmetic problems; and (2) the program asynchronously evolved with the proposed mechanism can be completed with mostly the same or shorter execution steps than the program evolved by the conventional GP.