What if the World Were Different? Gradient-Based Exploration for New Optimal Policies.
Rui SilvaFrancisco S. MeloManuela VelosoPublished in: GCAI (2018)
Keyphrases
- optimal policy
- markov decision processes
- decision problems
- reinforcement learning
- finite horizon
- dynamic programming
- state space
- long run
- finite state
- multistage
- infinite horizon
- average reward
- sufficient conditions
- dynamic programming algorithms
- state dependent
- average reward reinforcement learning
- initial state
- markov decision process
- serial inventory systems
- control policies
- policy iteration
- markov decision problems
- bayesian reinforcement learning
- action selection
- average cost
- reinforcement learning algorithms
- data mining