Methods for continuous optimization: active-set techniques, decomposition schemes and derivative-free approaches
Descrizione
Andrea Cristofari è risultato vincitore della procedura selettiva per n.1 posto di Ricercatore a tempo determinato tipologia A, SC 01/A6 - SSD MAT/09 - Dipartimento di ingegneria informatica automatica e gestionale Antonio Ruberti, Bando n. 2/2021 R TDA, Prot. n. 2734/2021.
In ottemperanza ai requisiti previsti dal bando, lunedì 21 marzo alle ore 15:45, Andrea Cristofari illustrerà le sue attività di ricerca svolte e in corso di svolgimento in un seminario pubblico.
Title: Methods for continuous optimization: active-set techniques, decomposition schemes and derivative-free approaches
Abstract: In this talk, I will present the main results of my research activity in the field of continuous optimization. In particular, active-set techniques are first described for several classes of constrained problems, combined with second-order and first-order search directions. Then, a decomposition scheme for problems with one linear equality constraint is presented and a derivative-free approach is described for structured optimization problems. The application of the above algorithms for machine learning and data science problems is discussed.
Short bio: Andrea Cristofari received the M.Sc. degree in Management Engineering (summa cum laude) and the Ph.D. degree in Automatic Control and Operations Research (with honors) from Sapienza University of Rome in 2013 and 2017, respectively. From 2016 to 2017, he was Postdoctoral Fellow at Sapienza University of Rome (Department of Computer, Control and Management Engineering "Antonio Ruberti"). From 2017 to 2019, he was Postdoctoral Fellow at University of Padua (Department of Mathematics "Tullio Levi-Civita"), where he is a fixed-term Researcher from 2019. His interests include algorithms for constrained and unconstrained problems of continuous optimization, especially active-set methods, decomposition methods and derivative-free methods, with a focus on large-scale problems and application in machine learning and data science.