Sparse Matrix Feature Selection in Multi-label Learning

High-dimensional data are commonly met in multi-label learning, and dimensionality reduction is an important and challenging work. In this paper, we propose sparse matrix feature selection to reduce data dimension in multi-label learning. First, the feature selection problem is formalized by sparse matrix. Second, an sparse matrix feature selection algorithm is proposed. Third, four feature selection are compared with the proposed methods and parameter optimization analysis is also provide. Experiments reported the proposed algorithms outperform the other methods in most cases of tested datasets.