Hands-On Neuroevolution with Python

The substrate builder function

The ES-HyperNEAT method runs the neuroevolution process, which includes the evolution of the CPPN genomes along with the evolution of the substrate configuration. However, even though the substrate is evolving during evolution, it is incredibly beneficial to start with an appropriate initial substrate configuration. This configuration should correspond to the geometry of the problem space.

For the retina experiment, the appropriate substrate configuration is created as follows:

1. First, we create the configuration of the input layer of the substrate. As you may remember from the The initial substrate configuration section, the eight nodes of the input layer are placed within the XZ plane, which is orthogonal ...