Abstract | ||
---|---|---|
We present a model-free reinforcement learning method for partially observable Markov decision problems. Our method estimates a likelihood gradient by sampling directly in parameter space, which leads to lower variance gradient estimates than those obtained by policy gradient methods such as REINFORCE. For several complex control tasks, including robust standing with a humanoid robot, we show that our method outperforms well-known algorithms from the fields of policy gradients, finite difference methods and population based heuristics. We also provide a detailed analysis of the differences between our method and the other algorithms. |
Year | DOI | Venue |
---|---|---|
2008 | 10.1007/978-3-540-87536-9_40 | ICANN (1) |
Keywords | Field | DocType |
detailed analysis,complex control task,policy gradient method,likelihood gradient,policy gradients,policy gradient,humanoid robot,observable markov decision problem,model-free reinforcement,variance gradient estimate,finite difference method,parameter-based exploration,reinforcement learning,parameter space,gradient method | Population,Decision problem,Mathematical optimization,Computer science,Markov chain,Heuristics,Artificial intelligence,Sampling (statistics),Finite difference method,Machine learning,Reinforcement learning,Humanoid robot | Conference |
Volume | ISSN | Citations |
5163 | 0302-9743 | 22 |
PageRank | References | Authors |
1.13 | 9 | 6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Frank Sehnke | 1 | 527 | 39.18 |
Christian Osendorfer | 2 | 125 | 13.24 |
Thomas Rückstieß | 3 | 112 | 20.66 |
Graves, Alex | 4 | 8572 | 405.10 |
Jan Peters | 5 | 3553 | 264.28 |
Jürgen Schmidhuber | 6 | 17836 | 1238.63 |