Rethinking Out-of-Distribution Detection for Reinforcement Learning: Advancing Methods for Evaluation and Detection.
Linas NasvytisKai SandbrinkJakob N. FoersterTim FranzmeyerChristian Schröder de WittPublished in: AAMAS (2024)
Keyphrases
- reinforcement learning
- detection algorithm
- automatic detection
- evaluation methods
- detection accuracy
- detection method
- significant improvement
- benchmark datasets
- preprocessing
- state space
- object detection
- gold standard
- false positives
- evaluation criteria
- machine learning algorithms
- model free
- optimal control
- evaluation method
- learning algorithm
- face detection
- dynamic programming
- bayesian networks