Biomorph: learning about neural and genetic algorithms in an artificial word

Biomorph is a Python (version 2.6) program for learning about various "biomorphic" artificial intelligence algorithms. It was used to teach the artificial intelligence course Neural and Genetic Approaches to Artificial Intelligence (Computer Science B553) at Indiana University in Spring 2009. The program runs in two modes, one implementing (and displaying) a world in which agents interact with and adapt to their environment, one implementing (and displaying) a neural network as it responds according to a particular learning algorithm.

Five compressed versions of the program are provided here:

• biomorph_rl.tgz, which contains parameters for reinforcement learning (Q learning)
• biomorph_ga.tgz, which contains parameters for genetic algorithms
• biomorph_som.tgz, which contains parameters for competitive learning and self-organizing maps
• biomorph_hop.tgz, which contains parameters for Hopfield networks
• biomorph_srn.tgz, which contains parameters for for feed-forward and simple recurrent neural-network learning

There is currently only minimal documentation for the program. If you have questions about its use, please contact Michael Gasser (gasser @ cs . indiana . edu).