GPU Evaluates a Population of a Million Running for a Thousand Generations

This page summarises the animation (20Mb) (YouTube video) of species of programs evolving, forming demes, spreading and becoming extinct. Each of the 1000 populations of a million programs was evaluated on an nVidia GeForce 8800 GTX GPU

Smaller movies can be found: YouTube video, quality animation. Genetic v. phenotypic film (MUN copy)

This movie is designed to show the variation in Amino acids used in the population of classifiers. (Description of Nuclear v non-Nuclear proteins task). There are 20 amino acids available to the programs. A program can use zero, or more. It can also use an Amino acid more than once. In which case the additional ones have less impact on the colours. The number of Amino acids of each type are grouped into a 20 dimensional space. The colour of each program is given by the angle between its 20 dimensional vector and that of the program in generation 1000 eventually chosen to be the best. These angles are plotted across the bottom in the lower histogram. (This program is halfway across right at the top)

Generation 0

The initial generation is made of randomly chosen programs. The log distribution below shows random program vary widely in their use of Amino acids.

Generation 10

The best of each pair is copied over the weaker program. This leads to an immediate visual roughening of the texture as many adjacent pixels now have similar colours.

Generations 20-100

Selection tends to reduced diversity in the population as areas of similar programs start to form and expand. Since selection, crossover and mutation are limited to adjacent pixels and these are rotated through the four cardinal directions (up, down, left, right).

The histograms reflects the slow movement towards the good program by tending to move to the left.

Generations 100-1000

The areas of similar programs tend to grow at the expense of other species.

W.Langdon xx.essex.ac.uk 6 June 2007 (last revised 7 Jun 2013)