Enhancing Explainability of Deep Reinforcement Learning Through Selective Layer-Wise Relevance Propagation.
Tobias HuberDominik SchillerElisabeth AndréPublished in: KI (2019)
Keyphrases
- reinforcement learning
- multiple layers
- function approximation
- machine learning
- information retrieval
- reinforcement learning algorithms
- model free
- application layer
- multi layer
- relative importance
- stochastic approximation
- state space
- relevance feedback
- temporal difference
- wave propagation
- test collection
- dynamic programming
- learning process
- pairwise
- feature selection
- learning algorithm