Conditions on Preference Relations that Guarantee the Existence of Optimal Policies.
Jonathan Colaço CarrPrakash PanangadenDoina PrecupPublished in: AISTATS (2024)
Keyphrases
- optimal policy
- preference relations
- decision problems
- sufficient conditions
- markov decision processes
- stationary policies
- finite horizon
- long run
- state space
- dynamic programming
- multistage
- infinite horizon
- reinforcement learning
- finite state
- average reward
- pairwise comparisons
- decision theory
- policy iteration
- desirable properties
- multi attribute
- dynamic programming algorithms
- partial order
- markov decision process
- average reward reinforcement learning
- semi markov decision processes
- serial inventory systems
- markov decision problems
- sample path
- lost sales
- initial state
- multiple agents