The Role of Lookahead and Approximate Policy Evaluation in Policy Iteration with Linear Value Function Approximation.
Anna WinnickiJoseph LubarsMichael LivesayR. SrikantPublished in: CoRR (2021)
Keyphrases
- policy evaluation
- policy iteration
- reinforcement learning problems
- markov decision processes
- least squares
- reinforcement learning
- model free
- temporal difference
- optimal policy
- fixed point
- monte carlo
- sample path
- markov games
- markov decision process
- variance reduction
- linear programming
- function approximation
- semi parametric
- finite state
- average reward
- convergence rate
- state space
- reinforcement learning algorithms
- markov decision problems
- neural network
- optimal control
- cost function
- average cost
- evaluation function
- optical flow
- machine learning