This paper introduces the correntropy-based semi-supervised multi-view deep autoencoder-like NMF (CSDANMF) framework, and provides comprehensive algorithmic evaluations, including a formal robustness analysis on corrupted datasets and a computational complexity analysis.
Abstract
Deep autoencoder-like nonnegative matrix factorization (DANMF) has emerged as a powerful dimensionality reduction paradigm, gaining significant traction in multi-view clustering (MVC) applications. Although existing DANMF-based MVC frameworks demonstrate competitive performance, they remain inherently susceptible to noise contamination and often fail to substantially improve clustering outcomes by effectively exploiting sparse supervisory information. To address these limitations, this paper introduces the correntropy-based semi-supervised multi-view deep autoencoder-like NMF (CSDANMF) framework for advanced multi-view clustering tasks. Compared to conventional DANMF-based MVC approaches, the proposed CSDANMF method introduces two distinct innovations: (1) CSDANMF substitutes the traditional linear Frobenius norm with a non-linear, localized similarity metric—specifically, maximum correntropy—as the loss function, thereby significantly enhancing the model’s robustness against outliers and noise. (2) CSDANMF leverages sparse label information to construct initial pairwise constraints and subsequently deploys a constraint propagation algorithm (CPA) to diffuse these supervisory signals across the data manifold, thereby maximizing the utility of limited prior knowledge to guide the clustering process. Furthermore, we provide comprehensive algorithmic evaluations, including a formal robustness analysis on corrupted datasets and a computational complexity analysis. Extensive experimental results across six benchmark nonnegative multi-view datasets demonstrate that CSDANMF consistently outperforms six state-of-the-art MVC methods, validating its efficacy and superior clustering performance.
Recently, multi-view learning techniques based on non-negative matrix factorization (NMF) have demonstrated remarkable efficacy in clustering multi-source data. However, existing methods often overlook the intrinsic similarity between views, and conventional NMF struggles to consistently yield sufficiently sparse solutions. To address these limitations, this paper proposes a novel NMF-based multi-view clustering approach. First, a pairwise co-regularization mechanism is introduced to capture cross-view structural correlations by measuring the similarity between view-specific coefficient matrices. Second, by imposing logarithmic norm sparsity constraints on the coefficient matrix, a pairwise co-regularized NMF model with sparsity constraints is constructed. An efficient iterative update algorithm is derived for the proposed model, and its convergence is rigorously guaranteed in theory. Extensive experiments on eight benchmark datasets validate the effectiveness of the proposed algorithm and its superior sparse representation capability. Furthermore, as a practical engineering optimization, a stochastic acceleration strategy is incorporated to expedite convergence. Empirical results show that this strategy reduces the runtime by approximately 48% on average while preserving clustering performance. Future work will focus on the theoretical underpinnings of this stochastic strategy and its applicability to broader scenarios.
Multi-view subspace clustering has progressed significantly by using deep neural networks to handle nonlinear data representations. A recent advancement, the Multi-view Self-Expressive Subspace Clustering (MSESC) network, achieves markedly higher computational efficiency by substituting the traditional self-expression layer with a deep metric learning approach. Nevertheless, MSESC still suffers from two notable limitations: it fails to adequately capture the high-order geometric structures inherent in multi-view data, and it lacks effective guidance from the underlying clustering distribution. To overcome these shortcomings, we propose a novel framework termed Multi-View Graph Regularized Deep Metric Subspace Clustering (MVGR-DMSC). The proposed method introduces two key components into MSESC to enhance the discriminability of representations. First, a dual-order graph regularization module is devised to maintain both first-order and second-order manifold structures, thereby allowing the model to capture more complex local geometric relationships. Second, an adaptive view-weighted deep clustering module is incorporated, which employs the Kullback–Leibler divergence to guide representation learning while dynamically adjusting the contributions of different views. Through evaluations on five benchmark datasets, we show that MVGR-DMSC consistently yields better results than several state-of-the-art approaches, including the direct baseline MSESC, in both accuracy and robustness.
Pengpeng Luo, Ming Yang, Chong Peng et al.· PLoS ONE· 0 citations
Multi-view clustering aims to integrate complementary information from diverse views; however, effectively improving clustering performance with limited supervision remains a significant challenge. Existing semi-supervised methods often fail to fully exploit sparse labeled data or lack effective mechanisms to propagate supervisory constraints across views. To address these limitations, we propose a novel semi-supervised multi-view clustering method based on concept factorization, termed DSCF. Specifically, DSCF establishes a unified framework that incorporates pointwise constraints to enforce clustering consistency for labeled samples, while a hypergraph-based pairwise constraint propagation strategy is employed to transfer supervisory information to unlabeled data. Furthermore, a cross-view iterative update scheme is designed to effectively align and fuse constraint information among different views. Extensive experiments on eight benchmark multi-view datasets demonstrate that DSCF consistently outperforms several relevant multi-view clustering methods, validating its effectiveness in leveraging limited labeled information to enhance clustering performance.
Bangding Yin, Xiaojun Yang· Digital Signal and Computer...· 0 citations
