Warm-up Free Policy Optimization: Improved Regret in Linear Markov Decision Processes.
Asaf CasselAviv RosenbergPublished in: CoRR (2024)
Keyphrases
- markov decision processes
- optimal policy
- total reward
- reward function
- expected reward
- policy iteration
- average reward
- finite horizon
- markov decision process
- infinite horizon
- state and action spaces
- action space
- state space
- finite state
- reinforcement learning
- average cost
- decision processes
- partially observable
- reachability analysis
- dynamic programming
- reinforcement learning algorithms
- policy iteration algorithm
- factored mdps
- markov decision problems
- decision problems
- decision theoretic planning
- planning under uncertainty
- policy evaluation
- model based reinforcement learning
- transition matrices
- continuous state spaces
- long run
- discounted reward
- multistage
- control policies
- stationary policies
- temporal difference
- sufficient conditions
- partially observable markov decision processes