Selective Breeding of Cellular Automata using Monochrome and Colourised Games of Life
The Game of Life is a universal computer, capable of computing anything a Turing machine can compute. A program in Life is written in a two-dimensional grid space, by setting the colour of each square in the grid to either white or black. With more than 50 years of experience, Life programmers have created a large collection of two-dimensional configurations of black and white squares. Each configuration of squares serves as a module that can perform a useful computation. In this article, we propose an alternative approach to the Game of Life. Instead of programming Life, we use an evolutionary algorithm to breed Life. We use a three-part approach, with (1) a population of small seed patterns, (2) a single large target pattern, and (3) an evolutionary algorithm. The population of seed patterns is initially random. A seed is randomly selected from the population and allowed to grow, and then it is assigned a fitness score, based on how well it matches with the large target pattern. The evolutionary algorithm makes small, random changes to the seed population, selecting seeds that increase the fitness of the match with the target pattern, and removing seeds that have low fitness. We demonstrate some success with this approach, using the Game of Life (monochrome black and white) and the Immigration Game (colourised white, red, and blue).
Keywords: cellular automata, evolutionary algorithms, Game of Life, breeding automata, colourised Life.