Transition Constrained Bayesian Optimization via Markov Decision Processes.
Jose Pablo FolchCalvin TsayRobert M. LeeBehrang ShafeiWeronika OrmaniecAndreas KrauseMark van der WilkRuth MisenerMojmír MutnýPublished in: CoRR (2024)
Keyphrases
- markov decision processes
- finite state
- reinforcement learning
- state space
- optimal policy
- dynamic programming
- policy iteration
- decision theoretic planning
- reachability analysis
- transition matrices
- finite horizon
- average reward
- planning under uncertainty
- risk sensitive
- factored mdps
- markov decision process
- action space
- model based reinforcement learning
- efficient optimization
- decision processes
- action sets
- state and action spaces
- stochastic shortest path
- average cost
- reinforcement learning algorithms
- reward function
- infinite horizon
- state abstraction
- partially observable markov decision processes
- discounted reward
- semi markov decision processes
- posterior probability
- objective function