MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts.
Weixin LiangJames ZouPublished in: CoRR (2022)
Keyphrases
- training dataset
- synthetic datasets
- training and testing data
- text classification tasks
- benchmark datasets
- pascal voc
- training and test data
- high dimensional datasets
- training data
- test data
- massive datasets
- ground truth labels
- uci datasets
- text classification
- training phase
- image dataset
- online learning
- bag of words
- probability distribution
- training process
- training examples
- contextual information
- object detection
- data distribution
- conflict resolution
- object recognition
- text classifiers
- training set
- test set
- linear svm
- training algorithm
- restricted boltzmann machine
- supervised learning
- photo collections
- class imbalanced data
- real life
- random variables
- outlier detection
- data mining