On Q-learning Convergence for Non-Markov Decision Processes.
Sultan Javed MajeedMarcus HutterPublished in: IJCAI (2018)
Keyphrases
- markov decision processes
- stochastic shortest path
- optimal policy
- policy iteration
- reinforcement learning algorithms
- state space
- reinforcement learning
- finite state
- dynamic programming
- markov games
- discounted reward
- reward function
- finite horizon
- markov decision problems
- decision theoretic planning
- average cost
- markov decision process
- infinite horizon
- discount factor
- reachability analysis
- action space
- transition matrices
- continuous state spaces
- partially observable
- convergence rate
- decision problems
- average reward
- decision processes
- stationary policies
- factored mdps
- stochastic games
- planning under uncertainty
- least squares
- model free
- convergence speed
- fixed point
- model based reinforcement learning
- real time dynamic programming
- cooperative