Solving Deep Memory POMDPs with Recurrent Policy Gradients.
Daan WierstraAlexander FörsterJan PetersJürgen SchmidhuberPublished in: ICANN (1) (2007)
Keyphrases
- markov decision problems
- partially observable markov decision processes
- optimal policy
- reinforcement learning
- policy iteration algorithm
- sequential decision making problems
- markov decision processes
- policy gradient methods
- policy gradient
- partially observable
- state space
- policy search
- utility function
- neural network
- distributed constraint optimization
- control policies
- state dependent
- policy iteration
- markov decision process
- main memory
- memory usage
- limited memory
- decision processes
- function approximation
- continuous state
- linear programming
- actor critic
- partially observable markov decision process
- decision theoretic planning
- recurrent neural networks
- dynamic programming
- action selection