MixLab: An Informative Semi-supervised Method for Multi-label Classification

Abstract

Multi-label classification is an intensively studied topic in data analysis. In spite of the considerable improvements, recent deep learning-based methods overlook the existence of unlabeled data, which consumes too much time on instance annotation. To circumvent this difficulty, semi-supervised multi-label classification aims to exploit the readily-available unlabeled data to help build multi-label classification model. To make full use of labeled and unlabeled data, this paper propose a novel approach named MixLab, encourages the model classifications to be accurate with label-correlated information and consistency regularization. It utilizes label correlations to enhance predicted labels for augmented unlabeled instances as targets and regularizes predictions to be consistent with this targets. We empirically validate the effectiveness of our framework by extensive experiments on four real datasets of textual content.