Multiply Accelerated Value Iteration for NonSymmetric Affine Fixed Point Problems and Application to Markov Decision Processes.
Marianne AkianStéphane GaubertZheng QuOmar SaadiPublished in: SIAM J. Matrix Anal. Appl. (2022)
Keyphrases
- markov decision processes
- fixed point
- policy iteration
- stochastic shortest path
- state space
- floating point
- optimal policy
- planning under uncertainty
- action space
- markov decision process
- reinforcement learning
- dynamic programming
- factored mdps
- average reward
- transition matrices
- decision theoretic planning
- real time dynamic programming
- infinite horizon
- finite state
- markov decision problems
- sample path
- partially observable
- finite horizon
- sufficient conditions
- probabilistic planning
- reinforcement learning algorithms
- risk sensitive
- partially observable markov decision processes
- policy evaluation
- higher order
- decision processes
- decision problems
- total cost