Tetris, Predator/Prey, and More

This is the version of the software used to produce results for publications associated with the 2017 Genetic and Evolutionary Computation Conference (GECCO 2017). Specifically, an implementation of HyperNEAT has been added, which was used to evolve state-evaluaters for Tetris. Full support for cooperative coevolution was also added to evolve agents in a toroidal predator/prey domain. The publications associated with these upgrades are:

Lauren E. Gillespie, Gabriela R. Gonzalez, and Jacob Schrum (2017). Comparing Direct and Indirect Encodings Using Both Raw and Hand-Designed Features in Tetris, Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2017).

Alex Rollins, and Jacob Schrum (2017). Balancing Selection Pressures, Multiple Objectives, and Neural Modularity to Coevolve Cooperative Agent Behavior, Companion to the Proceedings of the Conference on Genetic and Evolutionary Computation (GECCO Companion 2017). (movies, software).

Visit http://people.southwestern.edu/~schrum2/ for more details.

Isolated tasks

This version of MM-NEAT adds experiments in isolated versions of the Ms. Pac-Man task.