ACRE: Actor-Critic with Reward-Preserving Exploration.
Athanasios Ch. KapoutsisDimitrios I. KoutrasChristos D. KorkasElias B. KosmatopoulosPublished in: Neural Comput. Appl. (2023)
Keyphrases
- actor critic
- policy gradient
- reinforcement learning
- average reward
- optimal control
- gradient method
- temporal difference
- approximate dynamic programming
- action selection
- function approximation
- reinforcement learning algorithms
- optimal policy
- markov decision processes
- neuro fuzzy
- long run
- machine learning
- approximation methods
- policy iteration
- reward function
- single agent
- model free
- variance reduction
- state space
- convergence rate
- linear program
- monte carlo
- dynamic programming
- markov chain