ANJI (Another NEAT Java Implementation) Built on top of existing OpenSource projects, ANJI is an implementation of NEAT (Neuro-Evolution of Augmenting Topologies), an algorithm for evolving artificial neural networks developed by Kenneth Stanley, working in the Neural Networks Research Group at the University of Texas at Austin.

NEAT evolves both the connection weights and architecture of neural networks by starting with minimal topologies and adding innovations each generation through crossover and three types of mutations (changing the weight of an existing connection, adding a new connection, and adding a new neuron). For more information about NEAT, please click on the NEAT logo above to visit the NEAT Users Page.

Much of the evolutionary computing aspects are implemented using JGAP (Java Genetic Algorithms Package).

To learn more about JGAP, please click on the logo above.

To download the latest version of ANJI or access documentation please follow the links on the left.

Also, please feel free to contact us with any questions.

ANJI 2.0 Released! (8/20/05)