no code implementations • NeurIPS 2015 • Igor Mordatch, Kendall Lowrey, Galen Andrew, Zoran Popovic, Emanuel V. Todorov
We present a method for training recurrent neural networks to act as near-optimal feedback controllers.