Partially View-aligned Clustering (PVC) addresses the challenge of partial view alignment in multi-view learning by leveraging complementary and consistent information. While existing PVC methods show promise, most rely on distance-based strategies that are sensitive to view-specific details and noise, limiting their robustness. In this work, we propose a novel view alignment strategy that reformulates the alignment task as an anomaly detection problem. Rather than learning a view-alignment matrix that enforces strict one-to-one correspondences across views, we adopt a progressive approach to identify well-aligned samples. Specifically, we sample subsets of data by generating random view combinations from unaligned samples and propose an anomaly combination detection module to evaluate the alignment consistency of these combinations. In addition, our progressive training framework alternates between updating model parameters and selecting high-confidence view combinations for subsequent optimization. By reformulating view alignment as an anomaly detection task, our approach provides a more robust and effective solution to partial view alignment. Experiments on benchmark datasets demonstrate that our method outperforms state-of-the-art approaches in the PVC problem.

Identifying cancer subtypes is crucial for understanding disease progression. With advancements in high-throughput experimental technology, leveraging multiple types of omic data for subtype identification has become feasible. Various integrative cancer subtyping methods present a promising computational approach for identifying cancer subtypes from heterogeneous datasets. While existing integrative cancer subtyping methods have shown promising results in this task, efficiently integrating and clustering multi-omics datasets remains challenging due to high noise levels in omics data, which hinder accurate relationship capture among samples. To overcome this challenge, we propose a new deep multi-view subspace clustering model that introduces a self-reinforced learning strategy. This strategy iteratively enhances the quality of self-representation, crucial for capturing relationships among samples and for clustering. Specifically, during model training, our method is capable of learning a highly reliable self-representation by leveraging a good neighbor learning approach. This capability enables us to capture more accurate and robust relationships among samples. Subsequently, with the assistance of this highly reliable self-representation, we further develop a learnable view-graph fusion approach, which enables us to learn an accurate consensus for clustering and guides the overall model learning process. Additionally, we introduce a local graph-guided learning mechanism based on an initial graph learned from raw data. This mechanism helps prevent the model from converging to suboptimal solutions, thereby avoiding unsatisfactory and unstable results. Experimental results demonstrate that our method outperforms several state-of-the-art methods, verify the effectiveness of our approach in cancer subtype identification task.

In recent years, methods based on heterogeneous graph neural networks (HGNNs) have been widely used for embedding heterogeneous graphs (HGs) due to their ability to effectively encode the rich information from HGs into low-dimensional node embeddings. Existing HGNNs focus on neighbor aggregation and semantic fusion while neglecting the HG structure and learning paradigms. However, the original HG data might lack node features, which existing models may not effectively account for. Additionally, exclusively relying on a single supervised learning approach may only partially leverage the invariant information in graph data. To address these challenges, we introduce the Contrastive Auxiliary Learning Model for Heterogeneous Graphs (CALHG). This model combines edge perturbation and graph diffusion to enhance graph data, allowing it to capture the inherent structural information within heterogeneous graphs fully. Additionally, we employ a category-guided multi-view contrastive learning approach, which does not rely on positive and negative samples for model training, enabling us to capture the intrinsic invariances in heterogeneous graph data. Extensive experiments and analyses on five benchmark datasets without node features and three benchmark datasets with node features demonstrate the effectiveness and efficiency of our novel method compared with several state-of-the-art methods.

Multi-view clustering (MVC), which integrates information from multiple views to enhance performance, has garnered increasing attention in recent years. Partially View-aligned Clustering (PVC), which is a particularly critical aspect of this process, requires a thorough exploration of complementary and consistent information under conditions of partial view alignment. However, most existing PVC methods primarily focus on semantic consistency, employing semantic consistency features for both view alignment and clustering tasks. These methods neglect the effects of noise and complementary information across multiple views and the suitability of these features for clustering. To address these limitations, our approach aims to leverage three distinct types of consistency to extract semantic consistency features and clustering consistency features, which are specifically designed for view alignment and clustering tasks, respectively. By omitting the reconstruction process, we mitigate the adverse effects of mutual information and noise on view alignment. Specifically, we first exploit the structural consistency of similarity graphs across different views to guide feature extraction in view-specific autoencoders. This process produces structural consistency features that are both cluster-discriminative and structurally coherent. Subsequently, two separate multilayer perceptrons (MLPs) are trained via contrastive learning to extract semantic consistency features and clustering consistency features from the structural features. These features are optimized for their respective tasks. Ultimately, a self-paced style view alignment strategy is used to iteratively re-align the data based on semantic and clustering consistency while the model is optimized via the re-aligned data. Extensive experiments on multiple real-world benchmark datasets demonstrate that our method outperforms the state-of-the-art multi-view approaches, highlighting its effectiveness in tackling the challenges of PVC. The code is available at https://github.com/kongyiH/TCLPVC.

Partially View-aligned Clustering (PVC) presents a challenge as it requires a comprehensive exploration of complementary and consistent information in the presence of partial alignment of view data. Existing PVC methods typically learn view correspondence based on latent features that are expected to contain common semantic information. However, latent features obtained from heterogeneous spaces, along with the enforcement of alignment into the same feature dimension, can introduce cross-view discrepancies. In particular, partially view-aligned data lacks sufficient shared correspondences for the critical common semantic feature learning, resulting in inaccuracies in establishing meaningful correspondences between latent features across different views. While feature representations may differ across views, instance relationships within each view could potentially encode consistent common semantics across views. Motivated by this, our aim is to learn view correspondence based on graph distribution metrics that capture semantic view-invariant instance relationships. To achieve this, we utilize similarity graphs to depict instance relationships and learn view correspondence by aligning semantic similarity graphs through optimal transport with graph distribution. This facilitates the precise learning of view alignments, even in the presence of heterogeneous view-specific feature distortions. Furthermore, leveraging well-established cross-view correspondence, we introduce a cross-view contrastive learning to learn semantic features by exploiting consistency information. The resulting meaningful semantic features effectively isolate shared latent patterns, avoiding the inclusion of irrelevant private information. We conduct extensive experiments on several real datasets, demonstrating the effectiveness of our proposed method for the PVC task.

Clustering of the single-cell RNA-seq (scRNA-seq) transcriptome profiles is able to identify cell types, which is beneficial to improve the understanding of disease progression. However, in practice, the single-cell expression data often contains a significant number of missing values as a result of technical variability. Missing data is a critical challenge in scRNA-seq clustering analysis since the unknown value does not reflect the underlying true expression level and makes it difficult to discovering cell types by applying clustering algorithms directly. Various approaches have been developed to overcome missing data issue in scRNA-seq clustering. Most of them recover missing expression values by borrowing observed data from similar cells or synthesizing data via generative adversarial networks. Such that the biologically meaningful cluster structure has not been sufficiently exploited. In this work, we introduce ColImpute, a collaborative structure-preserved missing data imputation approach for the scRNA-seq clustering. Specifically, a cluster structure-preserved imputation module and a subspace clustering module, which respectively perform missing data imputation and cell subtypes identification, are integrated into a unified optimization framework to train the two networks in a collaborative manner. Consequently, the clustering module effectively contributes cluster-structure information to guide the trainning process of the missing data imputation module. Simultaneously, the cluster structure-preserved imputation module reciprocally enhances the performance of the clustering module by generating more precise recovered samples. Promising experimental results show that the proposed method is effective for both the data imputation and the cell types identification.

Deep subspace clustering has achieved remarkable performance in unsupervised clustering tasks. The self-supervised approach is further introduced to learn more discriminative representation for enhancing clustering performance. Despite the significant improvement of clustering performance by exploiting self-supervision information, these approaches heavily depend on the high quality of pseudo-label from the current clustering result and this will inevitably degrade the clustering performance when the obtained pseudo-labels are incorrect. To solve this issue, we develop a robust self-supervised deep subspace clustering approach by exploiting the reliable self-supervised information during training. The proposed method is involved in two key steps: a diffusion processing step is developed to improve self-expressiveness matrix such that more accurate clustering result (pseudo-labels) can be obtained. More importantly, we further propose to estimate and exploit the reliability of the assigned pseudo-label for each sample to alleviate the negative impact of incorrect pseudo-labels, such that the unreliable self-supervision can be further alleviated. Experimental studies on several benchmark datasets validate the effectiveness of our approach in terms of refining the self-supervised information. The source code of the proposed method is available at the https://github.com/stuljj/RSDSC.git.

Deep subspace clustering methods have attracted extensive attention due to the great improvement in both representation ability and precision of non-linear data. However, the rich information which is contained in the self-expression matrix is unexplored since existing approaches use the self-expression matrix only as a tool for learning inter-sample relationships and clustering. In addition, such models treat outlier and noise points equally with other points, which inevitably degrades the clustering performance. To overcome these issues, we develop a progressive deep subspace clustering approach by extracting delayed fitting probabilities from the module and then use the probabilities to defer the fitting of unreliable points. Specifically, we calculate the probabilities that each sample lies in each subspace based on the results of the self-expression matrix and spectral clustering, and then estimate the reliability of the cluster assignment of each sample as delayed fitting probability to reweight the loss of each sample. Experiments on five benchmark datasets validate the effectiveness of the proposed method.