Abstract In multi-label learning, label-specific feature learning can effectively avoid some ineffectual features that interfere with the classification performance of the model. However, most of the existing label-specific feature learning algorithms improve the performance of the model for classification by constraining the solution space through label correlation. The non-equilibrium of the label distribution not only leads to some spurious correlations mixed in with the calculated label correlations but also diminishes the performance of the classification model. Causal learning can improve the classification performance and robustness of the model by capturing real causal relationships from limited data. Based on this, this paper proposes a causality-driven intra-class non-equilibrium label-specific features learning, named CNSF. Firstly, the causal relationship between the labels is learned by the Peter-Clark algorithm. Secondly, the label density of all instances is calculated by the intra-class non-equilibrium method, which is used to relieve the non-equilibrium distribution of original labels. Then, the correlation of the density matrix is calculated using cosine similarity and combined with causality to construct the causal density correlation matrix, to solve the problem of spurious correlation mixed in the label correlation obtained by traditional methods. Finally, the causal density correlation matrix is used to induce label-specific feature learning. Compared with eight state-of-the-art multi-label algorithms on thirteen datasets, the experimental results prove the reasonability and effectiveness of the algorithms in this paper.

Abstract Multi-view multi-label (MVML) learning is a framework for solving the problem of associating a single instance with a set of class labels in the presence of multiple types of data features. The extraction of shared features among multiple views for label prediction is a common MVML learning method. However, previous approaches assumed that the number and association degree of shared features were the same across views. In fact, they differ in the number and degree of association. The above assumption can lead to a poor communicability of the views. Therefore, this paper proposes an MVML learning method based on the inconsistent shared features extracted by the graph attention model. The first step is to extract the shared and private features of multiple views. Next, the graph attention mechanism is adopted to learn the association degree of shared features of different views and calculate the adjacency matrix and attention coefficient. The number of associations is determined by taking the obtained adjacency matrix as a mask matrix, while the association degree of shared features is measured by the attention weight matrix. Finally, the new shared features are obtained for multi-label prediction. We conducted experiments on seven MVML datasets to compare the proposed algorithm with seven advanced algorithms. The experimental results demonstrate the advantages of our algorithm.