Mathematics Colloquia and Seminars

Return to Colloquia & Seminar listing

"A fundamental challenge in data science is the selection of discriminative features from a relatively small collection of sample pairs. A major difficulty is often that the relevant variables only arise as hidden factors in the actual raw data vectors.

One example are mass spectrometry data of the human proteome, where the desired
molecular concentrations of proteins are intrinsically encoded by means of
Gaussian-shaped peaks.

In this talk, we will develop a mathematical framework including in particular non-linear
observations, arbitrary convex signal structures as well as strictly convex loss functions,
which allows us to show that successful feature selection is still possible when the
applied estimator does not have any knowledge of the underlying data representation and
only takes the raw samples as input. Guarantees of such type are especially appealing for
practical purposes, since in many applications even standard methods, e.g., the Lasso
or logistic regression, yield surprisingly good outcomes."

Contact Jesus De Loera for further information