WEIGHTED SUPERPOSITION ENSEMBLE MULTIPLE LABEL CLASSIFICATION BASED ON SPARSE REGULARIZATION
In order to fully exploit the correlation of paired labels and the relationship between classifier weight and classifier selection,a weighted superposition ensemble multiple label classification method based on sparse regularization is proposed.A sparse regularized weighted superposition ensemble model was proposed to facilitate the selection of multiple label classifiers and the construction of ensemble members.The classifier weight and label correlation were used to improve the classification performance.An optimization algorithm based on accelerated proximal gradient and block coordinate descent technique was proposed to obtain the optimal solution effectively.Experimental results on several data sets show that the proposed method can effectively achieve high precision multiple label classification.
国家科技期刊平台
NSTL
万方数据
