Abstract | ||
---|---|---|
Most of the reinforcement learning (RL) algorithmsassume that the learning processesof embedded agents can be formulated asMarkov Decision Processes (MDPs). However,the assumption is not valid for many realisticproblems. Therefore, research on RLtechniques for non-Markovian environmentsis gaining more attention recently. We havedeveloped a Bayesian approach to RL in nonMarkovianenvironments, in which the environmentis modeled as a history tree model,a stochastic model with ... |
Year | Venue | Keywords |
---|---|---|
1997 | ICML | bayesian approach,non-markovian environments,model learning,reinforcement learning,stochastic model |
Field | DocType | ISBN |
Variable-order Bayesian network,Markov process,Pattern recognition,Computer science,Wake-sleep algorithm,Artificial intelligence,Graphical model,Machine learning,Model learning,Learning classifier system,Bayesian probability | Conference | 1-55860-486-3 |
Citations | PageRank | References |
3 | 0.41 | 1 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Nobuo Suematsu | 1 | 54 | 8.99 |
Akira Hayashi | 2 | 51 | 9.08 |
Shigang Li | 3 | 282 | 43.13 |