The Optimal Reward Baseline for Gradient-Based Reinforcement Learning.
Lex WeaverNigel TaoPublished in: UAI (2001)
Keyphrases
- reinforcement learning
- dynamic programming
- optimal control
- function approximation
- control policy
- average reward
- optimal solution
- state space
- total reward
- model free
- supervised learning
- machine learning
- optimal policy
- markov decision processes
- learning process
- reinforcement learning algorithms
- learning capabilities
- lower bound
- relative improvement
- worst case
- long run
- temporal difference
- learning algorithm
- eligibility traces