Encog NEAT and HyperNEAT Documentation

Encog contains support for three closely related neural network technologies: NEAT, CPPN and HyperNEAT. Kenneth Stanley’s EPLEX group at the University of Central Florida conducts extensive research for all three technologies. Information about their current research can be found at the following URL:

http://eplex.cs.ucf.edu/

NeuroEvolution of Augmenting Topologies (NEAT) is an algorithm that evolves neural network structures with genetic algorithms. The compositional pattern-producing network (CPPN) is a type of evolved neural network that can create other structures, such as images or other neural networks. Hypercube-based NEAT, or HyperNEAT, a type of CPPN, also evolves other neural networks. Once HyperNEAT train the networks, they can easily handle much higher resolutions of their dimensions.