Hands-On Neuroevolution with Python

The simple genetic algorithm

The simple genetic algorithm that's used in our experiment evolves the population of N individuals over generations of evolution. As we mentioned previously, each individual genome encodes the vector of the trainable ANN parameters. Also, in each generation, we select the top T individuals to become the parents for the next generation.

The process of producing the next generation is implemented as follows. For N-1 repetitions, we do the following:

1. The parent is selected uniformly at random and removed from the selection list.
2. The mutation is applied to the selected parent by applying additive Gaussian noise to the parameter vector that's encoded by the individual.
3. Next, we add the new organism to the list of ...