Host(s): Thor Bagge
In this talk I am going to give a basic introduction to Neuroevolution of Augmenting Topologies (NEAT).
The rather long and scary sounding name is a genetic algorithm for constructing neural networks to solve reinforcement learning tasks.
Just as neural networks are inspired by how our brains function, NEAT tries to imitate the biological evolution of genes with mutations in order to construct efficient networks.
We are going to start out with a basic introduction to neural networks, what they are and how they work.
Afterwards we are going to talk about NEAT and how the algorithm works.
Finally we look at an example of a NEAT implementation by Sethbling. This implementation attempts to learn how to effectively play Super Mario World.
EvaluationLogin to evaluate.