Evolving Novel Cellular Automaton Seeds Using Compositional Pattern Producing Networks (CPPN)
Joshuah Wolper, George Abraham · 2016
The aim of this study is to evolve novel seeds for John Conway's Game of Life cellular automaton (CA) with Compositional Pattern Producing Networks (CPPNs), a variation of artificial neural networks known to evolve organic patterns when used to process visual data. CPPNs were evolved using both objective search (implemented with NeuroEvolution of Augmenting Topologies) and novelty search, which focuses on finding novel solutions rather than objectively "fitter" solutions. Objective search quickly evolved game of life solutions that converged to trivial combinations of previously known solutions. However, novelty search produced non-trivial symmetries and complex high period oscillators such as the period 15 pentadecathlon. Regardless, neither approach evolved purely novel or undocumented seeds. Despite this failure, the complex evolved solutions demonstrate that CPPNs can serve as a powerful encoding for cellular automata seeds. As such, these results stand as the first baseline for further exploration into encoding cellular automata using CPPN.