Combined angular margin and cosine margin softmax loss for music classification based on spectrograms.
Jingxian LiLixin HanYang WangBaohua YuanXiaofeng YuanYi YangHong YanPublished in: Neural Comput. Appl. (2022)
Keyphrases
- hinge loss
- support vector
- loss function
- decision boundary
- training set
- soft margin
- feature vectors
- risk minimization
- classification accuracy
- pattern recognition
- support vector machine
- maximum margin
- potential functions
- decision trees
- feature extraction
- classification scheme
- classification method
- svm classifier
- class labels
- training samples
- nearest neighbor
- machine learning
- support vector machine svm
- image classification
- decision rules
- multi class
- binary classification
- classification models
- generalization error
- automatic classification
- multi layer perceptron
- music composition
- hyperplane
- pattern classification
- basis functions
- benchmark datasets
- text classification
- supervised learning
- preprocessing
- objective function
- feature selection