Label Distribution Learning

Abstract—Although multi-label learning can deal with many problems with label ambiguity, it does not fit some real appli-cations well where the overall distribution of the importance of the labels matters. This paper proposes a novel learning paradigm named label distribution learning (LDL) for such kind of applications. The label distribution covers a certain number of labels, representing the degree to which each label describes the instance. LDL is a more general learning framework which includes both single-label and multi-label learning as its special cases. This paper proposes six LDL algorithms in three ways: problem transformation, algorithm adaptation, and specialized algorithm design. In order to compare their performance, six evaluation measures are suggested for LDL algorithms, and the first batch of real-world label distribution datasets are prepared. Experimental results on the ten real-world datasets show clear advantage of the specialized algorithms, which indicates the importance of special design for the characteristics of the LDL problem. I