H IGH - DIMENSIONAL MULTI - VIEW CLUSTERING METHODS APREPRINT K. Jbilou Approximation and Numerical Analysis department ULCO University Dunkerque, France 62228 khalide.jbilou@univ-littoral.fr A. Ratnani Khwarizmi department Mohammed VI Polytechnic University Benguerir, Morocco 43150 ahmed.ratnani@um6p.ma A. Zahir Khwarizmi department Mohammed VI Polytechnic University Benguerir, Morocco 43150 alaeddine.zahir@um6p.ma March 16, 2023 ABSTRACT Multi-view clustering has been widely used in recent years in comparison to single-view clustering, for clear reasons, as it offers more insights into the data, which has brought with it some challenges, such as how to combine these views or features. Most of recent work in this field focuses mainly on tensor representation instead of treating the data as simple matrices. This permits to deal with the high-order correlation between the data which the based matrix approach struggles to capture. Accordingly, we will examine and compare these approaches, particularly in two categories, namely graph-based clustering and subspace-based clustering. We will conduct and report experiments of the main clustering methods over a benchmark datasets. Keywords Multi-view clustering · Graph based Multi-view clustering · Subspace based Multi-view clustering · Tensor learning · Representation learning 1 Introduction Clustering [Li et al., 2019, Yang and Wang, 2018, Zhao et al., 2017] is a fundamental topic of data mining as well as machine learning, especially when there are no labels for data objects. Clustering results are often used in subsequent applications, such as community detection, recommendation, and information retrieval using features. A feature is any individual measurable property or characteristic of an object, for example, an image is represented by different kind of features as colors, edges or texture, a document can be represented by different language. A video may be encoded in different amount of images and sound. The combine of these features are refereed as multi-view data, which it gave birth to a new paradigm. Although each view may have enough information by it’s own, it likely provides complementary information to the other views, this leaning paradigm is humanly nature, as we also tend to learn in a similar fashion, as in many real-life problems multi-view data arise naturally. The challenge in this new paradigm, is how to combine these features for a better result than the single view. In this paper, we focus on multi-view unsupervised learning and particularly, multi-view clustering. Note that we will focus on a more particular subject, that will be mentioned after giving the appropriate definition at the right time. One naive MVC technique is to use a single-view clustering algorithm on concatenated information acquired from several views. This strategy however, may fail if greater emphasis is placed on particular specific view than on others. As a result, multi view subspace clustering that learns common coefficient matrices, multi view k-means [Cai et al., 2013] and multi-kernel based MVC [Guo et al., 2014] have received increased attention. arXiv:2303.08582v1 [cs.LG] 14 Mar 2023