Machine Learning Group Seminars: "Stochastic Gradient Methods: Classical Guarantees under Lipschitz Smoothness and Their Limitations" - 2

We are excited to announce the relaunch of the Machine Learning Group Seminars at the Athens University of Economics and Business (AUEB), a joint initiative of the Departments of Statistics and Informatics!

The seminar series aims to bring together researchers and students interested in machine learning, statistics, optimization, and related areas. We are introducing a two-week format, designed to encourage a deeper exploration of selected topics. The first week will introduce the key ideas, concepts, and methodological background, while the second week will build on these foundations, potentially focusing on a specific research paper and its contributions.

First seminar: Vasilis Apidopoulos

We are delighted to kick off the renewed seminar series with Vasilis Apidopoulos, Assistant Professor at the Department of Informatics, AUEB.

Vasilis's research lies at the intersection of optimization and machine learning, with particular emphasis on first-order optimization algorithms, inertial and momentum methods, stochastic optimization, and their applications to machine learning and inverse problems. His work also explores the connections between optimization algorithms, continuous-time dynamical systems, and their convergence and stability properties.

Talk title:

Stochastic Gradient Methods: Classical Guarantees under Lipschitz Smoothness and Their Limitations

About the talk

In this introductory talk, Vasilis will present classical convergence results for (Stochastic) Gradient Descent under standard Lipschitz-smoothness assumptions, highlighting the theoretical foundations behind these widely used optimization methods.

The talk will also set the stage for the following week's seminar, which will explore mechanisms for taming stochastic optimization in settings where gradients may grow superlinearly. In particular, the discussion will focus on controlling stochastic updates and establishing almost-sure convergence to stationarity beyond the classical Lipschitz-smooth setting.

Seminar details:

Speaker: Vasilis Apidopoulos (Assistant Professor, Department of Informatics, AUEB)

Date: 19/11/2026 and 26/11/2026

Time: 14.00-15.00

Location: T107, Troias Building

We warmly invite faculty members, researchers, PhD/MSc/BSc students, and anyone interested in machine learning and optimization to join us. We look forward to seeing you there!

Ημερομηνία Εκδήλωσης: 
Thursday, November 26, 2026 - 14:00