Planning in entropy-regularized Markov decision processes and games.
Jean-Bastien GrillOmar Darwiche DominguesPierre MénardRémi MunosMichal ValkoPublished in: NeurIPS (2019)
Keyphrases
- markov decision processes
- action sets
- macro actions
- planning under uncertainty
- partially observable
- stochastic games
- decision theoretic planning
- reinforcement learning
- state space
- finite state
- optimal policy
- dynamic programming
- probabilistic planning
- policy iteration
- average reward
- average cost
- transition matrices
- planning problems
- heuristic search
- reachability analysis
- state and action spaces
- partially observable markov decision processes
- decision processes
- reinforcement learning algorithms
- model based reinforcement learning
- factored mdps
- initial state
- markov decision process
- objective function
- markov decision problems
- action space
- classical planning
- planning domains
- stationary policies
- blocks world
- finite horizon
- action selection
- infinite horizon
- stochastic shortest path