Discovering patterns and rules from data has been a major driver in scientific discoveries, highlighted by Kepler’s work about the rules governing the movements of planets. In many fields, progress often comes from a fruitful cross-fertilization between an empirical approach based on data and a theoretical approach based on the consequences of a few accepted fundamental principles. Myriads of quantitative data are currently available in many domains of application, like Astronomy, Earth Science, and Genomics, to name but a few. Data are often multidimensional, and algorithms for processing them are often polynomial in time with the dimension of the data (like computing an SVD of an \(n \times n\) matrix which is cubic with n). This makes some algorithms nonoperational for large dimensions and has led to the curse of dimensionality, coined by Bellman.