文档库

最新最全的文档下载
当前位置:文档库 > GENERATING MUSIC THROUGH GENETIC ALGORITHMS

GENERATING MUSIC THROUGH GENETIC ALGORITHMS

Science Research Programme 1996

GENERATING MUSIC THROUGH GENETIC ALGORITHMS

Lim Yu-Xi1, Lua Kim Teng2

ABSTRACT

Genetic algorithms attempt to embody Darwin’s Theory of Evolution in a series of computational steps, usually to derive optimal solutions to a given problem. This process is normally attributed to thinking and creativity but can be simplified in this case to a series of converging solutions. We have attempted to apply genetic algorithms to the composition of music, where the optimal solutions are the better sounding pieces. The main aim of this experiment was to derive an algorithm and find suitable values for its parameters. There was reasonable success in this experiment, which had a narrowed scope of classical music. One of the main problems encountered was defining what is good music.

INTRODUCTION

Intelligence and evolution are intimately linked. Intelligence is a natural part of higher lifeforms, but it can be broken down into simple steps and be simulated to a limited extent. One of the most daunting aspects of intelligence is creativity. Non-deterministic creativity is the spontaneous generation, derivation and selection of ideas. Interestingly, such a process can be found in another part of life – evolution.

Charles Darwin and Alfred Wallace presented their joint paper on the Theory of Evolution in 1858. It presented an idea which was highly controversial at the time when the word of the Roman Catholic Church was powerful. They proposed that the multitude of species that we observe today is the result of selective pressures against those that became extinct and for those that survive till today.

Such pressures work on the physical manifestations of the various traits of a given organism. Such traits are expressions of the organism’s genes, and the whole collection of genes is the organism’s chromosomes, or its genome.

Undesirable traits are gradually removed from the collection of genes, also know as the gene pool. This is because organisms having such undesirable traits have lower fitness, and are less likely to reproduce and propagate their genes. The opposite happens for desirable traits as they increase in frequency in the gene pool.

It is not always that undesirable traits become eliminated from the gene pool, not is it always that the gene pool will be dominated by a certain most desirable gene. Various combinations of genes can have different effects from the individuals from which it is derived. From here, it can be seen that it is likely that there will be several dominant chromosomes or genomes after a time. Each of these chromosomes will have desirable combinations of genes.

The genes do not just stagnate but undergo mutation at a constant rate. Random point mutations can change a specific gene to another. Other mutations reverse short portions of the

1 Science Research Programme student, Raffles Junior College, Singapore

2 Department of Information and Computational Science, National University of Singapore