Genetic Programming for Source Code Generation to Solve NP-hard Problems

Abstract

This paper describes the usage of genetic programming method for source code generationwith motivation to solve NP-hard problems. Described approach may be used in a wide range of modernapplications, whose working principle allows to apply optimization techniques. Proposed methodwas used to find a potential solution for achieving maximal score while playing a computer game called"Robocode tanks". The main principle of the experiment is based on applying classical evolution approacheson the selected problem in order to implement adaptive machine learning technique. During thetraining process of presented approach convergence starts and after several cycles of evolution, createdtank achieved significantly better final score compared to using a classic programming approach.This paper describes the usage of genetic programming method for source code generationwith motivation to solve NP-hard problems. Described approach may be used in a wide range of modernapplications, whose working principle allows to apply optimization techniques. Proposed methodwas used to find a potential solution for achieving maximal score while playing a computer game called"Robocode tanks". The main principle of the experiment is based on applying classical evolution approacheson the selected problem in order to implement adaptive machine learning technique. During thetraining process of presented approach convergence starts and after several cycles of evolution, createdtank achieved significantly better final score compared to using a classic programming approach.