Online Learning with Pairwise Loss Functions
Pairwise learning refers to a learning task which involves a loss function depending on pairs of examples, among which most notable ones include bipartite ranking, metric learning, minimum entropy error and AUC maximization. Online learning algorithms consider one instance each time to update the model sequentially, which make them amenable for streaming data analysis. However, most of such algorithms focused on the pointwise learning problems such as classification and regression.