On Solving MDPs With Large State Space: Exploitation of Policy Structures and Spectral Properties.
Libin LiuArpan ChattopadhyayUrbashi MitraPublished in: IEEE Trans. Commun. (2019)
Keyphrases
- markov decision problems
- state space
- optimal policy
- markov decision processes
- markov decision process
- reinforcement learning
- partially observable
- linear programming
- sequential decision making problems
- continuous state spaces
- semi markov decision processes
- action space
- reinforcement learning problems
- factored markov decision processes
- factored mdps
- policy iteration
- dynamic programming
- state and action spaces
- decision processes
- average reward
- reward function
- finite horizon
- policy search
- heuristic search
- reinforcement learning algorithms
- average cost
- decision theoretic planning
- transition probabilities
- utility function
- partially observable markov decision processes
- infinite horizon
- decision theoretic
- expected utility
- particle filter
- stochastic domains
- decision problems
- dynamical systems
- algebraic decision diagrams
- probabilistic planning
- state dependent
- initial state
- queueing networks
- state variables
- planning problems
- linear program
- markov chain
- search space
- control policies
- finite state
- sufficient conditions