Title: Clustering models for multi-view data Authors:  Paolo Giordani - Sapienza University of Rome (Italy) [presenting]
Abstract: It is common to encounter situations where variables are observed on a set of objects from multiple sources. In such cases, data are usually referred to as multi-view. Multi-view data offer a remarkably rich source of information, provided that suitable models are employed, as standard techniques often fall short. While models for multi-view data are frequently used for dimensionality reduction through components, clustering has garnered considerable attention in recent years. Typically, this involves clustering objects and compressing variables and sources through components. The clustering step is often performed using a hard partitioning approach, where objects are assigned exclusively to a single cluster. However, this can yield counterintuitive outcomes, particularly when objects exhibit characteristics shared across multiple clusters, which is frequently the case. Therefore, it is advantageous to partition objects using a fuzzy approach, allowing for soft membership degrees. After reviewing existing clustering models for multi-view data, fuzzy variants will be presented and examined through case studies.