The Pessimistic Limits of Margin-based Losses in Semi-supervised Learning.

Jesse H. Krijthe Marco Loog

Published in: CoRR (2016)

Keyphrases

semi supervised learning
unlabeled data
labeled data
semi supervised
semi supervised classification
supervised learning
co training
pairwise constraints

machine learning
unsupervised learning
training data
manifold regularization
learning problems
multi class
active learning
label propagation

labeled and unlabeled data
graph based semi supervised learning
semi supervised learning algorithms
improve the classification accuracy
regularization framework
generalization error
class labels

learning models
transfer learning
data points
domain adaptation
unlabeled samples
text classification
pattern recognition