PolyLoss: A Polynomial Expansion Perspective of Classification Loss Functions.
Zhaoqi LengMingxing TanChenxi LiuEkin Dogus CubukJay ShiShuyang ChengDragomir AnguelovPublished in: ICLR (2022)
Keyphrases
- loss function
- support vector
- hinge loss
- pairwise
- partially labeled data
- classification accuracy
- bayes rule
- machine learning
- squared error
- image classification
- support vector machine
- decision trees
- multi class
- generalization ability
- cross validation
- benchmark datasets
- risk minimization
- loss minimization
- feature selection
- logistic regression
- text classification
- semi supervised
- feature vectors
- class labels
- svm classifier
- support vector machine svm
- pattern classification
- learning to rank
- training samples
- reproducing kernel hilbert space
- training set
- feature space
- feature extraction
- convex loss functions