Genetic Algorithms and Machine Learning for Programmers

Did It Work?

If you start with the simplest possible rule, and only have one item in the population, not much changes. Your tournament only has one competitor. The mutation gives it a chance to improve, but it might not manage to turn on all of the cells. Here’s a typical run—it started with 12 live cells, and creeps up to 17 over time:

You need a larger population to allow your GA to try more rows. Experimentation suggested about 25 items works well. The algorithm can then manage to turn on all 32 bits within 20 epochs, like this:

Your choice ...