Constructing Intelligent Agents through Neuroevolution
The University of Texas at Austin
Neuroevolution, or training neural networks through genetic algorithms, can potentially solve difficult sequential decision tasks. Recurrent neural networks can be evolved to map sequences of states directly to optimal actions, which is a robust approach with continuous domains and with hidden states. In this talk, I will review recent advances in neuroevolution methods, and present several applications ranging from rocket control and autonomous vehicles to robotics and games. In particular, I will demonstrate the NERO machine learning game where the player trains a team of non-player characters interactively in real time.
Risto Miikkulainen is a Professor of Computer Sciences at the University of Texas at Austin. He received an M.S. in Engineering from the Helsinki University of Technology, Finland, in 1986, and a Ph.D. in Computer Science from UCLA in 1990. His recent research focuses on methods for evolving neural networks and applying these methods to game playing, robotics, and intelligent control. He is an author of over 200 articles on neuroevolution, connectionist natural language processing, and the computational neuroscience of the visual cortex. He is an editor of the Machine Learning Journal and Journal of Cognitive Systems Research.