PolyLoss: A Polynomial Expansion Perspective of Classification Loss Functions.
Zhaoqi LengMingxing TanChenxi LiuEkin Dogus CubukXiaojie ShiShuyang ChengDragomir AnguelovPublished in: CoRR (2022)
Keyphrases
- loss function
- support vector
- hinge loss
- bayes rule
- logistic regression
- classification accuracy
- partially labeled data
- pairwise
- loss minimization
- supervised learning
- image classification
- squared error
- pattern classification
- machine learning
- kernel classifiers
- update rules
- feature selection
- risk minimization
- support vector machine svm
- feature set
- support vector machine
- feature extraction
- benchmark datasets
- cross validation
- machine learning algorithms
- cost sensitive
- density estimation
- reproducing kernel hilbert space
- training samples
- nearest neighbor
- feature vectors
- decision trees