Robust Embedding Framework with Dynamic Hypergraph Fusion for Multi-label Classification

Abstract

Label embedding is an important family of multi-label classification algorithms which can jointly extract the information of all labels for better performance. However, few works have been done on multi-label embedding methods which can effectively deal with the interference of noisy data during training progress. Inspired by the low-rank hypothesis originating from label correlation, we propose a novel embedding based framework namely Robust Embedding Framework with Dynamic Hypergraph Fusion for multi-label classification (REFDHF), by which a latent space with strong anti-noise and predictive ability can be obtained. Meanwhile, a novel hypergraph fusion technology is designed to explore and utilize the complementary between feature space and label space to make the proposed REFDHF more robust. Furthermore, we conduct a deep extension for REFDHF, which is effectively applied to image annotation. Extensive experimental results on datasets with many labels demonstrate that our proposed approach is significantly better than the existing label embedding algorithms.