ICANN
2010
Springer
14 years 19 days ago
2010
Springer
We propose a new approach for reinforcement learning in problems with continuous actions. Actions are sampled by means of a diffusion tree, which generates samples in the continuou...