The use of a genetic algorithm to optimize the functional form of a multi-dimensional polynomial fit to experimental data

This paper begins with the optimisation of three test functions using a genetic algorithm and describes a statistical analysis on the effects of the choice of crossover technique, parent selection strategy and mutation. The paper then examines the use of a genetic algorithm to optimize the functional form of a polynomial fit to experimental data; the aim being to locate the global optimum of the data. Genetic programming has already been used to locate the functional form of a good fit to sets of data, but genetic programming is more complex than a genetic algorithm. This paper compares the genetic algorithm method with a particular genetic programming approach and shows that equally good results can be achieved using this simpler technique.

1 Introduction

The aim behind this work has been to locate the optimum of an objective function, based on a limited sampling of the function. The first part of the paper describes some initial work on the use of a simple genetic algorithm (GA) to optimise test functions chosen to be representative of the experimental data. The effects of the choice of crossover, parent selection, population size and mutation are analysed on these test functions.