Разреженные предварительно обученные представления

AbstractTraining data for text classification is often limited in practice, especially for applications with many output classes or involving many related classification problems. This means classifiers must generalize from limited …...