The BOS Webinar Series is back for its second season!
The first session will take place on Tuesday, September 1, 2026, at 4:00 PM CEST via Microsoft Teams.
Join the BOS mailing list to receive Microsoft Teams link.
This session will feature two 20-minute talks by Sanyou Mei and David Villacís, each followed by discussion.
Sanyou Mei
Title: First-order methods for bilevel optimization
Abstract: Bilevel optimization, also known as two-level optimization, is an important branch of mathematical optimization. It has found applications across various domains, including economics, logistics, supply chain, transportation, engineering design, and machine learning. In this talk, we will present first-order methods for solving a class of bilevel optimization problems using either single or sequential minimax optimization schemes. We will also discuss the first-order operation complexity of these methods and present preliminary numerical results to illustrate their performance.
Bio: Sanyou Mei is an Assistant Professor in the Department of Industrial Engineering and Decision Analytics (IEDA) at the Hong Kong University of Science and Technology. He received his Ph.D. degree from the Department of Industrial and Systems Engineering at University of Minnesota, and completed undergraduate studies in the School of the Gifted Young at University of Science and Technology of China with a B.S. degree in Mathematics in 2020.
David Villacís
Title: Beyond Strict Complementarity: A Variational Analysis Approach to Bilevel Hyperparameter Optimization with Sparse Regularization.
Abstract: Choosing one regularization weight per feature is a bilevel program with a nonsmooth lower level and thousands of upper-level variables. The standard hypergradient — implicit differentiation on the active support — assumes strict complementarity, and returns exactly zero at biactive coordinates, so the outer loop can never move them. Using an exact forward–backward reformulation and the closed-form coderivative of soft-thresholding, we build a hyper-subgradient whose only freedom is a sign per biactive coordinate, and a sign-consistent oracle to choose it. Its selection test is precisely a descent test, and it reduces exactly to the classical hypergradient under strict complementarity.
Bio: David Villacís is Assistant Professor of Applied Mathematics at Universidad Loyola Andalucía, Spain. He received his PhD in Applied Mathematics from Escuela Politécnica Nacional, Ecuador, in 2022, after an MSc in Computer Science from the University of Birmingham. Before joining Loyola he was a postdoctoral researcher at Universidad de O'Higgins and the Center for Mathematical Modeling in Chile. His work lies at the intersection of bilevel optimization, mathematical image processing, and machine learning, combining the variational analysis of nonsmooth problems with algorithm design and large-scale numerics.
We hope to see many of you online!
Best regards,
Yasmine Beck and Nagisa Sugishita