Realistic Evaluation of Test-Time Adaptation Algorithms: Unsupervised Hyperparameter Selection.
Sebastian CygertDamian SójkaTomasz TrzcinskiBartlomiej TwardowskiPublished in: CoRR (2024)
Keyphrases
- learning algorithm
- orders of magnitude
- times faster
- theoretical analysis
- real world
- realistic scenarios
- evaluation metrics
- significant improvement
- computational complexity
- supervised learning
- worst case
- machine learning algorithms
- lower bound
- computational efficiency
- recently developed
- selection algorithm
- evaluation method
- search algorithm
- gaussian processes
- evaluation methods
- neural network