A Proposed Dk-pc Algorithm For Code Bloat Control In A Tree-based Genetic Programming
This paper addresses the Genetic Programming (GP) issue of code bloat which is the uncontrolled growth of program codes without a commensurate improvement in the program fitness to solve a given problem. Code Bloat is a serious issue in GP as it consumes computer memory and processing time. Though, several reasons and solutions for code bloat control have been suggested in literature, yet no final solution has been found so far. Against this backdrop, we proposed the Delete lower and Keep higher fitness value Programs after Crossover (DKPC) algorithm which keeps the higher fitness value program and delete the lower value fitness value programs from memory. We tested the Boolean 6-multiplexer and Boolean 11-Multiplexer functions against the preparatory requirements using our proposed algorithm, and we got very impressive results; we observed that the algorithm was able to control bloat to a large degree. However, the algorithm performed better in the Boolean 11-multiplexer function than in the Boolean 6-multiolexer function. Both functions displayed almost the same behaviour; except that the Boolean 11-multiplexer exhibited higher performance result than the Boolean 6-multiplexer in terms of better program size reduction. To this extent, our algorithm performed better in bloat control, based on the benchmark problems used.
-
Rating
-
Date
October 2017 -
Size
1023.5KB -
Views
7,999 -
Categories