71 search results - page 1 / 15 » An Analysis of Direct Reinforcement Learning in Non-Markovia... |
ICML
1998
IEEE
14 years 5 months ago
1998
IEEE
NIPS
2001
13 years 6 months ago
2001
This paper presents reinforcement learning with a Long ShortTerm Memory recurrent neural network: RL-LSTM. Model-free RL-LSTM using Advantage learning and directed exploration can...
AGENTS
1999
Springer
13 years 9 months ago
1999
Springer
In this paper, we present a novel multi-agent learning paradigm called team-partitioned, opaque-transition reinforcement learning (TPOT-RL). TPOT-RL introduces the concept of usin...
JCP
2007
13 years 4 months ago
2007
Abstract— We describe a general method to transform a non-Markovian sequential decision problem into a supervised learning problem using a K-bestpaths algorithm. We consider an a...
ICML
2008
IEEE
14 years 5 months ago
2008
IEEE
Residual gradient (RG) was proposed as an alternative to TD(0) for policy evaluation when function approximation is used, but there exists little formal analysis comparing them ex...