Supervised learning as an inverse problem based on non-smooth loss function.
Soufiane LyaqiniMohamed QuafafouMourad NachaouiAbdelkrim ChakibPublished in: Knowl. Inf. Syst. (2020)
Keyphrases
- loss function
- supervised learning
- partially labeled data
- pairwise
- support vector
- empirical risk
- logistic regression
- training set
- training data
- active learning
- unsupervised learning
- semi supervised learning
- learning to rank
- machine learning
- learning algorithm
- boosting framework
- semi supervised
- reinforcement learning
- risk minimization
- solution path
- regularization term
- reproducing kernel hilbert space
- stochastic gradient descent
- class labels
- learning problems
- learning tasks
- convex loss functions
- generalization error