Multi-label feature selection can effectively solve the curse of dimensionality problem in multi-label learning. Existing multi-label feature selection methods mostly handle multi-label data without missing features. However, in practical applications, multi-label data with missing features exist widely, and most existing multi-label feature selection methods are not directly applicable. Therefore, we propose a feature selection method for multi-label data with missing features. First, we propose a method to extract implicit label information from the feature space to replenish the binary label information. Second, we learn the positive correlation between features to construct a feature correlation recovery matrix to recover missing features. Finally, we design a sparse model-based multi-label feature selection method for processing multi-label data with missing features and prove the convergence of this method. Comparative experiments with existing feature selection methods demonstrate the effectiveness of our method.

Multi-label feature selection is an effective approach to mitigate the high-dimensional feature problem in multi-label learning. Most existing multi-label feature selection methods either assume that the data is complete, or that either the features or the labels are incomplete. So far, there are few studies on multi-label data with missing features and labels. In many cases, missing features in instances of multi-label data often lead to missing labels, which is ignored by existing studies. We define this type of data as instance-dependent incomplete multi-label data. In this paper, we propose a feature selection method for instance-dependent incomplete multi-label data. Firstly, we use the positive correlations between features to reconstruct the feature space, thereby recovering missing values and enhancing non-missing values. Secondly, we use fuzzy tolerance relation to guide label recovery, and utilize fuzzy mutual implication granularity to impose structural constraint on the projection matrix. Thirdly, we achieve feature selection by eliminating the impact of incomplete instances and imposing sparse regularization on the projection matrix. Finally, we provide a convergent solution for the proposed feature selection framework. Comparative experiments with existing multi-label feature selection methods show that our method can perform effective feature selection on instance-dependent incomplete multi-label data.

In recent years, multi-view unsupervised feature selection has gained significant interest for its ability to efficiently handle multi-view datasets while offering better interpretability. However, most existing methods face the following challenges: First, the presence of noisy features in the data significantly impacts the process of learning accurate feature importance. Second, the selected features contain redundant information due to ignored redundancy between them. Third, graph structure learning is performed on all samples, resulting in large computational and space overheads, which is not conducive to expansion to large-scale data. To address these challenges, we propose a multi-view unsupervised feature selection method based on latent semantics and anchor graph learning. Specifically, this method designs a feature-weighted orthogonal regression and subspace learning framework to suppress noise interference in the consensus latent semantics discovery and anchor graph construction process, enhance the robustness of multi-view representation learning and reduce the computation of graph construction. Meanwhile, the proposed method employs explicit redundancy mitigation mechanisms that penalize discriminative weight allocation to highly correlated features. Furthermore, the proposed method unifies feature weighting, consensus latent semantics discovery, and adaptive graph learning within a multi-layer learning framework, enabling comprehensive feature importance evaluation through interactive learning between multiple layers. Finally, an efficient iterative algorithm is designed to solve the proposed model. The superiority of the proposed algorithm is demonstrated by comparing it with seven state-of-the-art algorithms on seven public multi-view datasets.

Processing large-scale data sets with limited labels has always been a difficult task in data mining. Facing this difficulty, two local feature selection algorithms, LARD and LRSD, have been proposed based on dependency degree, which can process partially labeled data sets and greatly improve the computational efficiency. However, it is very difficult for these algorithms to calculate large-scale data with millions of samples on a typical personal computer. Although the related family method is a more efficient approach than dependency degree, it cannot be used for partially labeled large-scale data. As a result, a local feature selection method based on related family is proposed to accelerate data processing in the paper. Experiments show that the proposed algorithm can run 405 times faster than LARD on partially labeled data sets and maintain high classification accuracy. In addition, this new algorithm can effectively process partially labeled large-scale data sets with 5,000,000 samples or 20,000 features on a typical personal computer.

On account of the rapid progress of zeroing neural network (ZNN) and the extensive use of fuzzy logic system (FLS), this article proposes an intelligent fuzzy robustness ZNN (IFR-ZNN) model and applies it to solving the time-variant Stein matrix equation (TVSME) problem. Be different from ZNN models before, the IFR-ZNN model uses a fuzzy parameter as the design parameter and adopts a first proposed improved nonlinear piecewise activation function. Particularly, the FLS that generates the fuzzy parameter utilizes an improved membership function of nonuniform distribution which can improve the adaptability and robustness of the IFR-ZNN model. Based on the above two optimizations, the proposed IFR-ZNN model possesses three significant advantages: (1) fixed-time convergence independent of initial states; (2) superior robustness to tolerate two kinds of noises simultaneously; and (3) better adaptiveness based on computational error. Besides, the upper bounds of fixed-time convergence of the IFR-ZNN model under noisy or non-noisy situations are calculated theoretically, and the stability as well as the excellent adaptability are analyzed in detail. Finally, simulation comparison results manifest the availability and meliority of the proposed IFR-ZNN model in solving the TVSME problem.