NEAT(NeuroEvolution of Augmenting Topologies) is a genetic algorithm that creates the artificial neural network. It works based on natural selection. As Darwin says, "Survival of the fittest." NEAT has shown greater promise in reinforcement learning tasks.
There are four steps in this algorithm:
The initial population Creates a population of birds(say 100).
Fitness function This function gives points to birds if they pass a certain threshold.
Collection of best Selection of best performers.
And mutate. Mutate them to make them even better.
Training Neural Network
The NEAT approach begins with a perceptron-like feed-forward network of only input neurons and output neurons. As evolution progresses through discrete steps, the network's topology's complexity may grow, either by inserting a new neuron into a connection path or by creating a new connection between neurons. Hence, it finds the optimal path.