Evolutionary sound synthesis

On the whole, evolutionary computation brings together a pool of algorithms for searching for one or more solutions to a problem in a vast space filled with possible solutions. In this context, the problem of finding the right s)mthesis parameters for producing a sound could be defined as a search for a point in a multi-dimensional space, whose co-ordinates correspond to the respective parameters of the s)mthesiser in question. 

Evolutionary algorithms require guidance to direct the evolution towards better areas of the search space. They receive this guidance by evaluating every solution in the population in order to determine its fitness. The fitness of a solution is a score based on how well the solution fulfils the problem in question. This score is calculated by means of a fitness function. 

The genetic algorithm, or GA, is perhaps the most popular evolutionary technique amongst musicans. The basic idea behind GAs first appeared in the early 1960s, but John Holland is often quoted as one of the pioneers of GAs for his work on modelling adaptive natural systems (Holland, 1975). 

GA theory uses jargon and a number of metaphors drawn from the biological sciences but readers should not be put off the principles behind the jargon and methaphors are fairly straightforward. 

In practice, GAs normally operate on a set of binary codes that represent the entities of the population. These operations involve three basic classes of processes recombination, mutation and selection. Whilst the recombination process leads to the exchange of information between a pair of codes, the mutation process alters the value of single bits in a code. Recombination produces offspring codes by combining the information contained in the codes of their parents. Depending on the form of the representation of the codes, two types of recombination can be applied real-valued recombination or binary-valued crossover. 

---