弱监督学习人工智能.pdfVIP

Background • • 3 Traditional Supervised Learning 4 Traditional Supervised Learning 5 • • • • • • Semi-Supervised Learning labeled examples Classifier unlabeled examples • • • 13 Teaching-to-Learn Learning-to-Teach (TLLT) Teacher 1: W(1) W(m) Teacher m: W(M ) Teacher M: Laplacian Welsch Regularization for Robust SSL Our model: Optimal solution: Acceleration: 15 Laplacian Welsch Regularization for Robust SSL Toy dataset Table. Average accuracy and CPU time