During the last several years, evidence has been accumulating that creative evolutionary systems based on genetic algorithms, genetic programming, and related methods can be surprisingly effective at deriving novel solutions to problems involving design, discovery and invention. We have been exploring this issue through a variety of computational experiments. Our most recent work has focused on designing antenna arrays, neural networks, and cellular automata models of self-replication using genetic programming. This current project is studying whether causal relations and cause and effect reasoning can be used, in part, to guide the evolutionary process in such systems. Specifically, we use causal relations to target where mutations and crossover points are applied to an individual's genetic representation.NSF Web Page
Jung J, Reggia J. Evolving an Autonomous Agent for non-Markovian Reinforcement Learning, Proc. Genetic and Evolutionary Computation Conference (GECCO), 2009.
Pan Z, Reggia J. Computational Discovery of Instructionless Self-Replicating Structures in Cellular Automata, Artificial Life, 16, 2010, 39-63.
Pan Z, Reggia J. Artificial Evolution of Self-Replicating and Problem-Solving Structures in Cellular Spaces, in Simulating Complex Systems by Cellular Automata, A. Hoekstra, J. Kroc and P. Sloot (eds.), Springer, 2010, 193-216.
Chabuk T, Reggia J. Causally-Guided Evolutionary Computation for Adapting Weights, Proc. Fourth International Conf. on Neural, Parallel and Scientific Computation, G. Ladde et al (eds.), Atlanta GA, 2010, 77-82.
Jacobs J, Reggia J. Evolving Musical Counterpoint, Proc. First Workshop on Evolutionary Music, 2011, in press.