Abstract | ||
---|---|---|
Q-learning and other linear dynamic learning algorithms are subject to Bellman's curse of dimensionality for any realistic learning problem. This paper introduces a framework for satisficing state abstraction -- one that reduces state dimensionality, improving convergence and reducing computational and memory resources -- by eliminating useless state dimensions. Statistical parameters that are dependent on the state and Q-values identify the relevance of a given state space to a task space and allow state elements that contribute least to task learning to be discarded. Empirical results of applying state abstraction to a canonical single-agent path planning task and to a more difficult multi-agent foraging problem demonstrate utility of the proposed methods in improving learning convergence and performance in resource-constrained learning problems. |
Year | DOI | Venue |
---|---|---|
2014 | 10.1109/ICMLA.2014.22 | ICMLA |
Keywords | Field | DocType |
q-learning,intelligent agent,state dimensionality,statistical parameter,reinforcement learning,statistical analysis,learning (artificial intelligence),resource-constrained learning problem,bellman curse of dimensionality,multi-agent systems,multiagent foraging problem,canonical single-agent path planning task,state abstraction,linear dynamic learning algorithm,complexity reduction,feature extraction,noise,learning artificial intelligence,indexes,vectors,convergence | Algorithmic learning theory,Semi-supervised learning,Stability (learning theory),Instance-based learning,Computer science,Theoretical computer science,Unsupervised learning,Artificial intelligence,Computational learning theory,Machine learning,Reinforcement learning,Learning classifier system | Conference |
Citations | PageRank | References |
0 | 0.34 | 11 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Cheng Zhao | 1 | 47 | 8.34 |
Laura E. Ray | 2 | 0 | 0.34 |