Region enhanced neural Q-learning for solving model-based POMDPs.
Marco A. WieringThijs KooiPublished in: IJCNN (2010)
Keyphrases
- reinforcement learning
- state space
- model free
- cooperative
- markov decision problems
- optimal policy
- multi agent
- sequential decision making problems
- learning algorithm
- stochastic shortest path
- neural network
- function approximation
- sequential decision problems
- partially observable
- partially observable markov decision processes
- dynamic programming
- reinforcement learning algorithms
- action selection
- image regions
- markov decision processes
- fully unsupervised
- stochastic approximation
- dynamic environments
- distributed constraint optimization
- point based value iteration
- solving problems
- exact solution
- associative memory