NEAT – Neuroevolution of augmenting topologies


NEAT (neuroevolution of augmenting topologies) is a project that combines Artificial Neural Networks and Genetic Algorithms.

NEAT has been developed for the first time in 2007 in Texas University and its innovation consists in the fact that the ANN modifies its weights and topology, miming the functioning of an organic NN.

This implementation will permit to implement a NEAT system easily and without knowing in details the functioning of this. This will allow to solve problems that requires an ANN (image recognition, clustering, finding hidden patterns…) quickly and using ANN with topologies built accordingly to the problem.




EXAMPLE: Pig Implementation



Using the NEAT is possible to define the behaviour of a group of pigs: their simple goal is to avoid falling in canyons, find a zone with trees and spend there their lives eating fruits.

Due to the simplicity of the task, the Artificial Neural Networks need just a few generations to get trained. Once obtained the behaviour we desire we can just extract the neural network of the individuals and apply it to the entire population.

Modeling the behaviour of a group of individual is just a silly example of what the NEAT can do


0 1 2 3 4