Adrien Taylor : Exploring every avenue of optimisation

He admits it himself: he likes to investigate there where it is not easy for him from the outset. And so, although he would have perhaps gravitated more naturally towards computer science, he decided to embark on engineering studies in applied mathematics. Then, particularly intrigued by the lectures of professors Moritz Diehl and Yurii Nesterov on optimisation, this was the direction he decided to take for his thesis in 2012 at the University of Louvain (UCL). From then onwards he has devoted himself to scientific optimisation, which offers him a vast area of exploration - with machine learningat the forefront. 

Building computer-assisted proofs

“During my thesis, I wanted to delve deeper into the points which had seemed the most mysterious to me during my studies. But, with my way of doing things, that meant using approaches that weren't necessarily standard”, he explains. In concrete terms, his work focuses on optimisation algorithms: “Nobody wants to try out all the methods and leave a computer running for two years without knowing if one of them will work. Building the proofs of their correct operation is traditionally done by hand and, in the end, it is often difficult to understand the complex path that has led to such a result. In my thesis, the idea was to be able to systematically produce these proofs using a computer.”

Prize-winning work 

This approach did not go unnoticed. In 2018, Adrien Taylor was a finalist in the very competitive Tucker Prize, sponsored by the international association Mathematical Optimization Society(MOS) to reward the best theses in the field of optimisation. He is also a winner of the ICTEAM Thesis Awardawarded by his university, UCL, whose main selection criteria were notably the nature and originality of the questions addressed and of the methodology used. The same year, his work was once again recognised by the IBM Innovation Award, awarded by the Belgian Fund for Scientific Research (FRS-FNRS). 

A post-doctorate to deepen and broaden horizons

After defending his thesis at the beginning of 2017, the adventure continued. Adrien Taylor turned to a post-doctorate degree in order to go further with the potential opened up by his work. He therefore decided to apply to join Inria and its Sierraproject team, which positions itself at the crossroads between the worlds of applied mathematics, statistics and computer science. “One of my primary motivations was to work with references in the scientific community on the subject of optimisation, such as Francis Bach, head of Sierra, or Alexandre d'Aspremont.”This choice also enabled him to open up his field of investigation further: “The thesis is, after all, rather solitary work. Which, for me, is very different with a post-doctorate. It allows me to cooperate much more with the different members of the team, but also to forge international collaborations. For example, this year I had the opportunity of working with the universities of Wisconsin-Madison and California (UCLA) in the United States, in Sweden with Lund University, but also with Dutch universities (Tilburg and Delft), and with a researcher from Google. It really is an opportunity to combine your skills and opinions with those of others."

And afterwards?

Adrien Taylor fully intends to put this window of opportunity - made possible by the post-doctorate - to good use afterwards: “What I like the most in the scientific world is this freedom to learn about different, very varied, fields. Moreover, that is why I decided to focus on optimisation, which can have applications in disciplines as different as mathematics, physics, medicine, computer science but also the economy, etc. My objective thereafter is to continue to explore this diversity.”

Adrien Taylor's career in a few dates

• 2006-2011:university studies in applied mathematics at the University of Louvain (UCL) and at KULeuven
• 2012-2017:thesis "Convex Interpolation and Performance Estimation of First-order Methods for Convex Optimization"at UCL under the supervision of François Glineur and Julien Hendrickx.
• 2017:post-doctorate at UCL and start of a post-doctorate at Inria – SIERRA project team, for a duration of two years
• 2018:finalist of the Mathematical Optimization Society(MOS) Tucker Prize, winner of the ICTEAM Thesis Awardpresented by UCL and of the IBM Innovation Award(FRS-FNRS)