Approximate dynamic programming with $(\min, +)$ linear function approximation for Markov decision processes.
Chandrashekar LakshminarayananShalabh BhatnagarPublished in: CoRR (2014)
Keyphrases
- approximate dynamic programming
- function approximation
- reinforcement learning
- markov decision processes
- policy iteration
- model free
- average cost
- function approximators
- factored mdps
- optimal policy
- temporal difference
- state space
- dynamic programming
- finite state
- reinforcement learning algorithms
- control policy
- policy evaluation
- actor critic
- machine learning
- learning algorithm
- average reward
- partially observable
- multi agent
- infinite horizon
- reward function
- optimal control
- decision processes
- action space
- action selection
- markov decision process
- real valued
- transfer learning
- multistage
- radial basis function
- supervised learning
- finite number
- neural network