Kernelized Q-Learning for Large-Scale, Potentially Continuous, Markov Decision Processes.
Isaac J. SledgeJosé C. PríncipePublished in: IJCNN (2018)
Keyphrases
- markov decision processes
- continuous state spaces
- state space
- optimal policy
- policy iteration
- reinforcement learning
- reinforcement learning algorithms
- action space
- markov games
- stochastic shortest path
- discounted reward
- dynamic programming
- reward function
- finite state
- factored mdps
- transition matrices
- markov decision process
- infinite horizon
- planning under uncertainty
- finite horizon
- partially observable
- discount factor
- reachability analysis
- average cost
- state action
- decision processes
- average reward
- state and action spaces
- policy evaluation
- state abstraction
- decision theoretic planning
- learning algorithm
- sufficient conditions
- model free
- action sets
- model based reinforcement learning
- multistage
- decision problems
- control problems
- continuous state
- markov decision problems
- hierarchical reinforcement learning
- long run
- machine learning
- markov chain
- initial state
- partially observable markov decision processes
- cooperative
- multi agent
- action selection