Tied-Augment: Controlling Representation Similarity Improves Data Augmentation.
Emirhan KurtulusZichao LiYann N. DauphinEkin Dogus CubukPublished in: CoRR (2023)
Keyphrases
- data sets
- training data
- raw data
- database
- complex data
- data quality
- similarity measure
- high quality
- synthetic data
- data collection
- image data
- small number
- data processing
- graph representation
- original data
- knowledge discovery
- data points
- end users
- neural network
- nearest neighbor
- computer systems
- distance function
- sensor data
- data distribution
- data analysis
- noisy data
- knowledge base
- data mining