Planning in Markov Decision Processes with Gap-Dependent Sample Complexity.
Anders JonssonEmilie KaufmannPierre MénardOmar Darwiche DominguesEdouard LeurentMichal ValkoPublished in: CoRR (2020)
Keyphrases
- markov decision processes
- sample complexity
- planning under uncertainty
- decision theoretic planning
- partially observable
- theoretical analysis
- state space
- learning problems
- upper bound
- reinforcement learning
- finite state
- special case
- learning algorithm
- generalization error
- active learning
- supervised learning
- policy iteration
- partially observable markov decision processes
- lower bound
- heuristic search
- optimal policy
- planning problems
- average reward
- probabilistic planning
- upper and lower bounds
- transition matrices
- dynamic programming
- sample size
- average cost
- training examples
- infinite horizon
- markov decision process
- learning tasks
- dynamical systems
- np hard
- semi supervised
- training data
- markov decision problems
- data mining