An Empirical Study on Distribution Shift Robustness From the Perspective of Pre-Training and Data Augmentation.
Ziquan LiuYi XuYuanhong XuQi QianHao LiRong JinXiangyang JiAntoni B. ChanPublished in: CoRR (2022)
Keyphrases
- data distribution
- data sets
- synthetic data
- database
- data processing
- data collection
- raw data
- statistical analysis
- data sources
- decision trees
- training examples
- data analysis
- data structure
- data points
- prior knowledge
- machine learning
- knowledge discovery
- training data
- training samples
- data acquisition
- neural network
- data quality
- historical data
- training dataset
- uniformly distributed
- labelled data