Well, Philip and I finally released the code for our opensource neuroevolution implementation, ANJI, last night.

We've been working on it part time for about a year and a half now, and we've used it to carry out a nice bit of preliminary research. We've authored a paper, "A Comparative Analysis of Simplification and Complexification in the Evolution of Neural Network Topologies", which we'll be presenting at the Genetic and Evolutionary Computation Conference (or GECCO) in Seattle this month. The talk schedule is here.

Anyway, it's satisfying to release a piece of opensource research software around the same time we've completed our first piece of original research.

If you're at all interested in this stuff, and are reasonably computer-savvy, download the software and have a go. There's no slick interface, so you'll have to run it from a command line. But if you like, you'll be able to replicate our Tic-Tac-Toe runs and evolve your own neural network players...again, if you're into that sort of thing.

And if anyone who reads this blog does download and play around with ANJI, we'd love to hear any feedback.