UAI
2001
13 years 9 months ago
2001
There exist a number of reinforcement learning algorithms which learn by climbing the gradient of expected reward. Their long-run convergence has been proved, even in partially ob...