About me
I am an Assistant Professor at the Institute for Mathematical and Computational Engineering of the Catholic University of Chile. I have also been a Visiting Researcher in the Algorithms Group at Google since July 2023.
My main research interests are the complexity theory of convex optimization (deterministic and stochastic). This research also relates importantly to some aspects of machine learning theory and private data analysis. I am also interested in understanding incentives and equilibria from game-theoretic perspectives.
I obtained an Engineering degree (Applied Math M.Sc. equivalent) from the Mathematical Engineering Department of Universidad de Chile, and a Ph.D. in Algorithms, Combinatorics and Optimization at Georgia Tech.
e-mail: hyperlink
My research is currently funded by:
FONDECYT Regular. Algorithmic Stability and Differential Privacy in Optimization, Machine Learning and Market Equilibria (PI) . 2021-24
INRIA Associate Teams. FOAM: First-Order Accelerated Methods for Machine Learning (co-PI with A. d'Aspremont). 2022-24
Anillo: Information and Computation in Market Design (Associate Researcher). 2022-24
Basal. CENIA: National Center for Artificial Intelligence (Young Researcher). 2022-32
Advising
Postdoc
Juan Pablo Contreras (Math, PUC)
Ph.D.
[2024-] Juan Pablo Flores (Math, PUC)
[2021-] Clément Lezane (Appl. Math, UTwente)
Master
[2021-2023] Tomás González (Eng, PUC) Currently Ph.D. in Machine Learning at CMU (USA)
[2021-22] Giorgio Micali (Appl. Math, UTwente). Currently Ph.D. in Statistics at U Twente (Netherlands)
[2019-22] Juan Pablo Flores (Math, PUC)
[2019-21] Patricio Ulloa (Eng, PUC)
[2019-20] Santiago Armstrong. Joint with Carlos Sing-Long (Eng, PUC)
[2018-20] Javiera Riffo. Joint with Jorge Vera (Eng, PUC). Currently M.Sc. in Statistics at TU Dortmund (Germany)
Visiting Students
Max Kramkimel: April-August 2023 (Ecole
Juan Pablo Contreras: May-July 2022 (UAI, Chile)
Nicolas Casabianca: April-August 2018 (ENS-Cachan, France)
Service
Conferences/Workshops Program Committees (w/proceedings)
NeurIPS 2024 (area chair), COLT 2024 (senior PC), LATIN 2024, OPT+ML Neurips Workshop 2023 (chair), COLT 2023 (senior PC), OPT+ML Neurips Workshop 2022, FORC 2022, GameNets 2021, COLT 2022
Conferences/Summer Schools Organization
Optimization, Learning and Games workshop 2023 (Roberto Cominetti's 60th birthday), XVI Discrete Mathematics Summer School 2021, Optimization and Learning workshop 2018
Regular Referee (for the past 5 years)
Mathematical Programming (meritorious service award 2022), NeurIPS, COLT, STOC
News
I was the chair of the OPT+ML workshop at NeurIPS 2023
During July-December 2023 I was a Visiting Researcher in the Discrete Algorithms Group at Google Research
Papers accepted in 2023:
Differentially Private Algorithms for the Stochastic Saddle Point Problem with Optimal Rates for the Strong Gap (with R. Bassily & M. Menart). COLT 2023
Faster Rates of Convergence to Stationary Points in Differentially Private Optimization (with R. Arora, R. Bassily, T. González, M. Menart & E. Ullah). ICML 2023
Optimal Algorithms for Differentially Private Stochastic Monotone Variational Inequalities and Saddle-Point Problems (with D. Boob). Mathematical Programming Series A, 2023In 2022 I was selected for the Mathematical Programming Meritorious Service Award
Recent/upcoming events:
19-12-'22: Talk in Workshop on Optimization and Algorithms (WOA) (Codegua, Chile)
05-10-'22: Talk in Learning and Optimization in Luminy (LOL) workshop (CIRM, Marseille)
23-09-'22: Talk at the Mathematics Colloquium (PUC-Chile)
08-09-'22: Joint talk (with J. Dunstan) at the IDIA Seminar (U. Chile)
08-08-'22: Plenary talk at ENIM 2022 (Santiago, Chile). The slides of my talk are here!
07-'22: (Cancelled) Talk in ICCOPT 2022 (Lehigh U, PA)
05-'22: Talk at the Journées Scientifiques INRIA Chile, Chile
04-'22: Talk at the Machine Learning Seminar, École Polytechnique, Paris, France
03-'22: Talk at the Data Privacy: Foundations and Applications Reunion (Simons Institute, Berkeley)
12-'21: Plenary talk at the Opt+ML workshop (NeurIPS 2021)