Title: A new biclustering method for functional data
Authors: Jacopo Di Iorio - Politecnico di Milano (Italy) [presenting]
Abstract: In recent years, thanks to the augmented possibilities in storing data, researchers started to deal with problems described by data having a huge number of features. It is the case of functional data, usually represented by a set of functions taking values into an infinite dimensional space. Another fundamental need, typical of data mining, is to study data by grouping them according to some measure of similarity: it is possible thanks to clustering techniques. Due to the fact that a large number of these algorithms cannot perform simultaneous and overlapping clustering on both the dimensions of the data, it has been proposed a new family of technique under the name of Biclustering or Co-Clustering. However, differently from clustering, there is not a large literature dedicated to these approaches suitable for functional data. The first attempts to deal with Biclustering for functional data are shown. A new possible technique to obtain biclusters in a functional setting is shown from both a theoretical and algorithmic points of view. Examples and real problem applications are analyzed and described in their results in order to highlight the importance of the introduction of these methods to the world of FDA.