Conditional generation model with dual-perspective feature fusion representation for multi-label classification

Xiaozhen Fu, Deyu Li, Erliang Yao, Yuhua Qian, Yang Li, Suge Wang

Abstract: Highlights•A deep conditional generative model is proposed for multi-label classification.•A variational inference framework and the variational objective are formulated.•A label-related fusion representation is generated from multiple perspectives.•Extensive experiments demonstrate the effectiveness and the superiority of FDML.