Rémi Le Priol

This is the website I used to communicate during my Ph.D (2017-2021). I no longer keep it up to date. Recent news are on LinkedIn.

About me

In my Ph.D. I explored ideas ranging from causality to physics and optimization, with the general purpose of making deep learning more robust. I was lucky to benefit from the supervision of Simon Lacoste-Julien and Yoshua Bengio. I also got the opportunity to work on these topics at Facebook AI Research with Léon Bottou. I enjoy science, especially when it relies on beautiful mathematics or when it is made accessible with appropriate illustrations, and I am striving to make this true in my own research. You can find my full list of publications on Google Scholar.

Short bio: I grew up and studied in Paris area, most notably at Ecole Polytechnique (X2013) and M2 MVA, before starting my Ph.D. in 2018. Details in my resume. Whether in or out of research, I enjoy truthfully relating to my fellow human beings, which often takes the form of acroyoga, or non-violent communication.

Some projects I care about

DualSnakes an interactive visualization I build on during Summer 2020. It illustrates projective duality and its connection to Fenchel duality, by displaying the dual of any curve drawn by the user.

DualityViz an interactive visualization I built during Spring 2020's lockdown. It aims at providing intuitions for convex duality, by showcasing the algebraic formulas on preset or user-provided examples, with a particular focus on the space-slope duality.

SDCA4CRF During the year 2017 I worked on an optimization algorithm for a structured prediction model called CRF. Along the way I found out a way to systematically improve the convergence speed of the algorithm SDCA by sampling data points adaptively.

bitstring-knitting a visual inspection of distances applied to bit strings.