The Geometry of Memoryless Stochastic Policy Optimization in Infinite-Horizon POMDPs.
Johannes MüllerGuido MontúfarPublished in: ICLR (2022)
Keyphrases
- infinite horizon
- optimal policy
- policy iteration algorithm
- partially observable
- markov decision processes
- finite horizon
- partially observable markov decision processes
- dec pomdps
- dynamic programming
- long run
- periodic review
- stochastic demand
- markov decision process
- reinforcement learning
- optimal control
- markov decision problems
- control policies
- production planning
- policy iteration
- state dependent
- decision problems
- single item
- state space
- average cost
- finite state
- stochastic programming
- average reward
- fixed cost
- lead time
- inventory policy
- optimal production
- holding cost
- multistage
- linear programming
- belief state
- multiple agents
- expected cost
- total reward
- machine learning