Averaged-DQN: Variance Reduction and Stabilization for Deep Reinforcement Learning.
Oron AnschelNir BaramNahum ShimkinPublished in: ICML (2017)
Keyphrases
- variance reduction
- reinforcement learning
- policy gradient
- gradient estimation
- sample size
- monte carlo
- random numbers
- bias variance decomposition
- function approximation
- reinforcement learning algorithms
- actor critic
- confidence intervals
- machine learning
- learning algorithm
- state space
- transfer learning
- dynamic programming
- importance sampling
- naive bayes classifier
- quasi monte carlo
- model selection