Multi-Label Learning Using Mathematical Programming

Hyunki Lim, Jae-Sung Lee, Dae-Won Kim

Abstract: We propose a new multi-label feature selection method that does not require the multi-label problem to be transformed into a single-label problem. Using quadratic programming, the proposed multi-label feature selection algorithm provides markedly better learning performance than conventional methods.

External IDs:dblp:journals/ieicet/LimLK15