Principled approaches and tools for the analysis and design of first-order optimization algorithms

Title

Principled approaches and tools for the analysis and design of first-order optimization algorithms

Speaker

Adrian Taylor - INRIA, École Normale Supérieure

Abstract

In this talk, we will present a high-level overview of systematic approaches for analyzing and designing (mostly first-order) optimization algorithms. The presentation will be illustrated through concrete examples, as the core principles underlying the approaches are already present in fundamental optimization schemes.

The methodology is implemented in the PEPit! package (https://pepit.readthedocs.io/), which enables users to apply the framework without the potential coding pain. This talk is based on joint work with great collaborators, who will be acknowledged during the presentation.

Bio

Adrien Taylor is a research scientist at Inria, working within the SIERRA team. He previously held a postdoctoral position in the same team from 2017 to 2019, collaborating with Francis Bach. He completed his PhD in mathematical engineering at Université catholique de Louvain, under the supervision of François Glineur and Julien Hendrickx, supported by a F.R.S.-FNRS FRIA scholarship. His research focuses on first-order optimization, numerical analysis, and aspects of control and machine learning. Committed to promoting reproducible and understandable science, his work often emphasizes transparency in both theory and practice. In recognition of his contributions, he was awarded an ERC Starting Grant in 2024 for the CASPER project, running from fall 2024 to fall 2029.

When

Friday June 6th, 12:00

Where

Room 322, UniGe DIBRIS/DIMA, Via Dodecaneso 35