Title: Robust and simple active learning schemes for Gaussian Processes and other regressors
Abstract: Many fields of science and engineering require the use of complex and computationally expensive models to understand the involved processes in the system of interest. Nevertheless, due to the high cost involved, the required study becomes a cumbersome process. In this talk, we describe different regression (or interpolation) procedures based on active learning algorithms, in order to construct cheap surrogate models of such costly complex systems. The techniques are then adaptive regression schemes where new data points are acquired sequentially. They are based on the optimization of suitable acquisition functions. We discuss the possible use of different acquisition functions. Specific computational details are remarked. The application to remote sensing problems is described. Finally, a description of how applying the same concepts to construct scalable regressors is also provided.
Luca Martino has obtained the MSc degree in Electronic Engineering at the Politecnico di Milano and his PhD inStatistical Signal Processing from Universidad Carlos III de Madrid (UC3M), Spain, in 2011. From 2013 to 2018 He worked as Assistant Professor and postdoctoral researcher at UC3M, at the University of Helsinki, at the Universidade de São Paulo (USP) and at the University of Valencia. Currently, he is Associate Professor at Universidad Rey Juan Carlos (URJC), Madrid, Spain. His research interests include Bayesian inference, Computational algorithms, Gaussian Processes and machine learning techniques.