.png){kind=icon}
.png){kind=icon}
Return
Cross-granularity hierarchy based on huffman coding for label-specific feature learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2025.132143.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Label-specific feature learning focuses on the specific attributes of individual labels themselves, which effectively improves the accuracy of multi-label classification. However, previous label-specific feature methods treat all labels as the same granularity hierarchy, and ignore the possible multi-granularity hierarchy among labels. In a way, exploiting the specific features of each label hierarchy can effectively reduce the complexity of label-specific feature learning. Moreover, coarse-grained predictions complement fine-grained feature representation, and fine-grained features improve the learning of coarse-grained classifiers. To improve the classification performance and reduce the complexity of multi-label learning, a novel hierarchy-specific features learning, dubbed CGH2C, is proposed in this paper. To the best of our knowledge, the method we present is the first to utilize the Huffman coding strategy for constructing the multi-granularity hierarchy among labels. Firstly, construct the Huffman tree through the Huffman coding theory to obtain the hierarchical structure of the labels. Then, the sparse linear model is used to obtain the local hierarchy-specific features. Finally, the extreme learning machine is used to predict the labels of each hierarchy. CGH2C achieves optimal performance over several benchmark multi-label datasets compared to existing methods. Experimental results and comprehensive model analysis demonstrate the effectiveness of the proposed method.