A novel Q-learning algorithm with function approximation for constrained Markov decision processes.
K. LakshmananShalabh BhatnagarPublished in: Allerton Conference (2012)
Keyphrases
- function approximation
- markov decision processes
- reinforcement learning
- learning algorithm
- reinforcement learning algorithms
- state space
- optimal policy
- policy iteration
- finite state
- temporal difference learning
- model free
- learning tasks
- learning problems
- machine learning
- temporal difference
- supervised learning
- control problems
- multi agent
- machine learning algorithms
- average reward
- function approximators
- partially observable
- policy evaluation
- learning agent
- average cost
- infinite horizon
- learning process
- active learning
- action space
- dynamic programming
- markov chain
- transfer learning
- reward function
- search algorithm
- learning rate
- markov decision process
- training data
- neural network
- continuous state