Reduced-GEP: Improving Gene Expression Programming by Gene Reduction

In traditional Gene Expression Programming (GEP), each chromosome is expressed and evaluated on the Expression Tree (ET). The ET-based expression and evaluation are computationally expensive and the intelligibility of the chromosome is low. In this paper, a highly efficient algorithm, Reduced-GEP, is proposed to solve these problems. First, the chromosome is reduced by Reduced-GEP. Second, chromosomes are evaluated directly on the reduced gene without being expressed them into ETs. In this way, the efficiency of the fitness evaluation is greatly improved. Moreover, the result of the evolution by Reduced-GEP is simplified and easier to be understood and explained. Extensive experiments demonstrate that Reduced-GEP algorithm is effective to calculate the fitness and reduce the chromosome.