Neuroevolution of augmenting topologies

Neuro Evolution of Augmenting Topologies (NEAT ) is the name of a genetic algorithm, the evolved artificial neural networks. It was developed in 2002 by Ken Stanley at the University of Texas at Austin. Due to its practical applicability of the algorithm is used in various fields of machine learning. Will be evolved, both the topology as well as the weights of the links in the neural network. The essential features of NEAT are:

• Assigning an identification number ( engl. innovation number) that allows the advantageous recombination of different topologies,
• The niche formation by restricting recombination to a relationship and
• The increasing diversity of the population with initial uniformity.

The extension allows the evolution HyperNEAT much larger networks by geometric structure of the given problem can be utilized (for example, controlling a plurality of legs).