Abstract:Multi-label learning is a subfield of machine learning that addresses the issue of each instance belonging to numerous class labels at the same time. However, in some real applications, we can only receive a partial set of labels for each instance due to the difficulty and high cost of labeling data. The vast majority of existing multi-label classification methods on missing labels rely on first- or second-order label correlation learning to fill in the original label space while building multi-label learning models with label-specific features; nevertheless, the single label correlation learning mechanism used in these methods is insufficient to maintain the consistency of the feature-label space. To address this issue, we propose the CLSML approach, which incorporates higher-order label correlation learning constraints in the classifier training model to complete missing labels while training the classifier. In addition, to improve the consistency of the feature-label space, we develop a two-stage second-order label correlation learning technique based on cosine similarity to further confine the label output. Furthermore, we employ the -norm regularizer to learn label-specific feature representations, followed by the -norm regularizer to constrain the row sparsity of the classification matrix and select label-common features. Experimental results comparing ten cutting-edge multi-label learning algorithms with missing labels on fourteen multi-label benchmark datasets demonstrate the effectiveness of our suggested approach.

Semi-supervised Multi-Label Learning with Missing Labels via Correlation Information

Learning Low-Rank Label Correlations for Multi-label Classification with Missing Labels

Weakly Supervised Multi-label Learning Using Prior Label Correlation Information

A Graph-based Semi-supervised Multi-label Learning Method Based on Label Correlation Consistency

Multi-Label Learning with Global and Local Label Correlation

Learning Common and Label-Specific Features for Multi-Label Classification With Missing Labels

Global and Local Attention-Based Multi-Label Learning with Missing Labels

Joint Label-Specific Features and Correlation Information for Multi-Label Learning.

Multi-label classification method based on correlation-constrained matrix factorization

Correlation-Induced Label Prior for Semi-Supervised Multi-Label Learning

Learning common and label-specific features for multi-Label classification with correlation information

Learning Instance Correlation Functions for Multilabel Classification

Learning Label-Specific Features for Multi-Label Classification with Missing Labels

Adaptive Collaborative Correlation Learning-based Semi-Supervised Multi-Label Feature Selection

A Pairwise Ranking System for Multi-Label Learning with Missing Labels

Beyond Missing: Weakly-Supervised Multi-Label Learning with Incomplete and Noisy Labels

Effective Semi-Supervised Learning Based On Local Correlation

Improving Multi-label Learning with Missing Labels by Structured Semantic Correlations

Joint Multi-Label Classification and Label Correlations with Missing Labels and Feature Selection

Multi-label classification with weak labels by learning label correlation and label regularization

Improving Multi-Label Classification with Missing Labels by Learning Label-Specific Features