What Are Effective Labels for Augmented Data? Improving Calibration and Robustness with AutoLabel.
Yao QinXuezhi WangBalaji LakshminarayananEd H. ChiAlex BeutelPublished in: SaTML (2023)
Keyphrases
- image data
- data sets
- data collection
- training data
- original data
- complex data
- raw data
- data sources
- high quality
- data points
- synthetic data
- probability distribution
- input data
- prior knowledge
- data quality
- end users
- data processing
- computer systems
- statistical analysis
- training examples
- high dimensional data
- data distribution
- data structure