The Brain in a Box An Encoding Scheme for Natural Neural Networks

摘要: To study the evolution of complex nervous systems through artificial development, an encoding scheme for modeling networks is needed that reflects intrinsic properties similiar to natural encodings. Like ge- netic code, a description language simulations should indirectly encode networks, be stable but adaptable and functions neural architectural design as well sin- gle neuron configurations. We propose indirect based on Compositional Pattern Produc- ing Networks (CPPNs) fulfill these needs. The uses CPPNs generate multidimensional patterns represent analog protein distributions in development organisms. These form template three-dimensional which dendrite- axon cones are placed space determine actual connections spiking network simulation.