Hi

Victor Sanches Portella

I am a postdoc researcher at the Institute of Mathematics and Statistics of the University of São Paulo (IME-USP) supervised by prof. Yoshiharu Kohayakawa. I obtained my PhD from the Department of Computer Science at UBC where I was advised by prof. Nick Harvey. In 2019 obtained a Master's degree in Computer Science from the Institute of Mathematics and Statistics of the University of São Paulo under the supervision of prof. Marcel K. de Carli Silva.

Research interests: I am broadly interested in numerical optimization, randomized algorithms, and differential privacy. Even more broadly I am interested in theoretical computer science and machine learning.

E-mail: victorsp [AT] ime [DOT] usp [DOT] br

PUBLICATIONS and PREPRINTS

Lower Bounds for Private Estimation of Gaussian Covariance Matrices under All Reasonable Parameter Regimes (In Submission, presented at TPDP 2024)
Authors: VSP, Nick Harvey. [arXiv]
On the Expected ℓ∞-norm of High-dimensional Martingales (Stochastic Processes and their Applications)
Authors (ABC ordering): Nick Harvey, Christopher Liaw, VSP. [PDF][SP&A Journal]
Searching for Optimal Per-Coordinate Step-sizes with Multidimensional Backtracking (NeurIPS 2023)
Authors: Frederik Künstner, VSP, Mark Schmidt, Nick Harvey. [arXiv] [NeurIPS 2023]
Continuous Prediction with Experts’ Advice. (JMLR)
Authors: VSP, Christopher Liaw, and Nick Harvey. [arXiv][JMLR]
Efficient and Optimal Fixed-Time Regret with Two Experts.(ALT 2022).
Authors (ABC ordering): Laura Greenstreet, Nick Harvey, VSP. [arXiv] [ALT 2022]
Regret Bounds without Lipschitz Continuity: Online Learning with Relative-Lipschitz Losses. (NeurIPS 2020).
Authors: Yihan Zhou*, VSP*, Mark Schmidt, and Nick Harvey. [arXiv] [NeurIPS 2020]
Online Mirror Descent and Dual Averaging: Keeping Pace in the Dynamic Case. (ICML 2020, Extended version in JMLR).
Authors: Huang Fang, Nick Harvey, VSP, Michael Friedlander. [arXiv] [ICML 2020] [JMLR]

THESES

Privacy, Experts, and Martingales: an Investigation on the Use of Analytical Tools.
Ph.D. Thesis, 2024. [webpage]
Online Convex Optimization: Learning, Duality, and Algorithms.
Master's Thesis, 2019. [webpage] [pdf]
Almost Linear Time Algorithms for Flows in Graphs.
Undergraduate Thesis, 2016. [webpage] [pdf]

MISCELLANEOUS

Google Scholar page.
Slides for some of my talks.
Resume (PDF)
I have recently made a video explainer on the HyperLogLog algorithm as part of the Summer of Math Exposition 1.
In 2018 I helped to port form Python 2 to Python 3 a framework for visualizing algorithms for computational geometry. You can find the port here (usage instructions are in portuguese). Prof. Cristina Gomes is the usual lecturer the the computational geometry course at the University of São Paulo and she might have more up to date versions of this tool.
Github page.