2A. From Data Analysis to Data Science

14:30 - 15:40, Aula 9

Organizer: Ndeye Niang

Chair: Ndeye Niang

Optimal Scaling: New Insights Into an Old Problem

Gilbert Saporta

Abstract: Processing qualitative variables with a very large number of categories in Machine Learning is an opportunity to revisit the theory of optimal scaling and its applications.

Click here to view the abstract.

Linear Approximation For Multivariate Categorical Functional Data Analysis

Cristian Preda and Quentin Grimonprez

Abstract: Multivariate categorical functional data is considered in the framework of dimension reduction through principal component analysis. Each statistical unit is represented by a vector of categorical functions observed on some time interval [0,T]. The computational algorithm for the optimal encoding functions and principal components is then adapted through a linear approximation for the encoding functions. A numerical study using Markov simulated data illustrates the accuracy of such an approximation.

Click here to view the abstract.

An overview of multi-view clustering approaches

Mohamed Nadif

Abstract: Various clustering models and algorithms derived from different perspectives are valuable for analyzing data collected from various sources or represented in multiple ways, each offering a unique insight into the data; see for instance [3, 1]. Therefore, when dealing with such data, many approaches can be used, such as factorization techniques [5], probabilistic methods [6, 4, 7], or even deep neural networks [2]. The effectiveness of clustering algorithms typically depends on how the data is represented, emphasizing the importance of learning a suitable data representation. Consequently, integrating these two tasks is a common strategy for exploring this type of data. This presentation will explore, discuss, and demonstrate the range of approaches employed, ranging from traditional to cutting-edge methods.

Click here to view the abstract.