Online Target Q-learning with Reverse Experience Replay: Efficiently finding the Optimal Policy for Linear MDPs.
Naman AgarwalSyomantak ChaudhuriPrateek JainDheeraj Mysore NagarajPraneeth NetrapalliPublished in: ICLR (2022)
Keyphrases
- optimal policy
- markov decision processes
- finite horizon
- decision problems
- state space
- reinforcement learning
- dynamic programming
- average reward
- infinite horizon
- multistage
- finite state
- policy iteration
- average cost
- long run
- bayesian reinforcement learning
- reward function
- discount factor
- markov decision problems
- markov decision process
- state dependent
- sufficient conditions
- reinforcement learning algorithms
- initial state
- decision processes
- partially observable markov decision processes
- monte carlo
- stochastic shortest path