ABrain’s documentation¶
This package implements the ES-HyperNEAT algorithms for the production of large-scale, regular Artificial Neural Networks. For a adequate overview of the related literature see the official homepage (http://eplex.cs.ucf.edu/ESHyperNEAT/).
Currently, the package can evolve both 2D and 3D ANNs and also exposes a generic CPPN e.g. for body/brain co-evolution. Illustrative examples are available under Usage and the full API is documented under its own section.
Notable missing features:
Crossover and historical markings (WIP)
Built-in NEAT evolution algorithms (WIP)
Hardware-accelerated ANN activation as a sparse-matrix, when relevant (WIP)