Multi-view clustering integrates multiple feature sets, which usually have a complementary relationship and can reveal distinct insights of data from different angles, to improve clustering performance. It remains challenging to productively utilize complementary information across multiple views since there is always noise in real data, and their features are highly redundant. Moreover, most existing multi-view clustering approaches only aimed at exploring the consistency of all views, but overlooked the local structure of each view. However, it is necessary to take the local structure of each view into consideration, because individual views generally present different geometric structures while admitting the same cluster structure. To ease the above issues, in this paper, a novel multi-view subspace clustering method is established by concurrently assigning weights for different features and capturing local information of data in view-specific self-representation feature spaces. In particular, common clustering assignment regularization is adopted to explore the consistency among multiple views. An alternating iteration algorithm based on the augmented Lagrangian multiplier is also developed for optimizing the associated objective. Clustering aims to organize reasonably unlabeled data to discover meaningful patterns. Clustering has an important role in many fields, including machine learning data mining and pattern recognition. Clustering methods can be mainly categorized into two groups: partitioning clustering and hierarchical clustering.  In particular, spectral clustering is a graph-based algorithm for partitioning arbitrarily shaped data structure into disjoint clusters. Numerous spectral clustering methods and their variants have been proposed, such as Ratio Cut, K-way Ratio Cut, Min Cut, Normalized Cut and Spectral Embedded Clustering. The clustering performance of all of these methods is largely dependent on the quality of the so-called similarity graph, which is learned according to the similarities between the corresponding data points. Recent works on spectral clustering-based subspace clustering have attracted considerable attention due to the promising performance in data clustering. Subspace clustering works on the assumption that data are drawn from a union of low-dimensional subspaces, i.e., every subspace is equivalent to a cluster. According to the self-expression of data and by imposing a properly chosen constraint on the representation coefficients, a representation matrix uncovering the intrinsic subspace structure of data can be obtained. For example, Elhamifar and Vida proposed sparse subspace clustering, which learns a graph by adaptively and flexibly selecting data points. Liu et al.  imposed a low-rank constraint on the representation matrix to capture the global structure of data. Dornaika and Weng  incorporated a manifold regularization term into SSC to capture the manifold structure. More recently, Peng et al. proposed a new deep model—Structured auto Encoder by preserving the global and local structure of subspace for clustering. Nevertheless, these approaches are mostly applied to group single-view data. In the era of Big Data, various data sources are represented by multiple distinct feature sets. Traditional clustering methods can be used to group multi-view data by simply concatenating all views into a monolithic one. However, compatible and complementary information across all views can be typically under-utilized. Recently, numerous multi-view clustering approaches have become available. Bickel and Scheffer generalized the conventional K-means and expectation–maximization clustering methods to the multi-view case.

Journal Homepage: https://obesity.imedpub.com/

Regards,
Catherine
Journal Co-Ordinator
Journal of Obesity and Eating DIsorders