NN
2010
Springer
13 years 10 months ago
2010
Springer
We present a model-free reinforcement learning method for partially observable Markov decision problems. Our method estimates a likelihood gradient by sampling directly in paramet...