SIFTER: Space-Efficient Value Iteration for Finite-Horizon MDPs.
Konstantinos SkitsasIoannis G. PapageorgiouMohammad Sadegh TalebiVasiliki KantereMichael N. KatehakisPanagiotis KarrasPublished in: Proc. VLDB Endow. (2022)
Keyphrases
- space efficient
- finite horizon
- markov decision processes
- optimal policy
- markov decision process
- infinite horizon
- optimal stopping
- data structure
- state space
- reinforcement learning
- average cost
- policy iteration
- data streams
- finite state
- sliding window
- single product
- dynamic programming
- b tree
- partially observable
- average reward
- action space
- stochastic shortest path
- control policies
- bloom filter
- factored mdps
- yield management
- decision problems
- machine learning
- markov decision problems
- range sum queries
- long run