Computational neuroscience · Machine learning

The geometry of
visual information.

Statistics. Symmetry. Information.

I study how visual systems, in brains and in machines, turn the structure of natural scenes into neural codes: which statistics they exploit, which transformations they respect, and what each neuron ends up telling us about the world.

Simulated OFF () and ON () ganglion-cell mosaics tracking a dark object. Move your cursor: the raster records from the cells under it.

Portrait of Simone Azeglio

About

Vision, in brains and machines.

I am a computational neuroscientist working between machine learning and vision neuroscience, currently a Guest Research Associate in Yukiyasu Kamitani's lab at Kyoto University.

I did my PhD at the Vision Institute (Sorbonne) and École Normale Supérieure with Ulisse Ferrari, Peter Neri, Olivier Marre and Matthew Chalk.

My research has two sides. I build models of visual neurons that carry the structure of natural scenes, their statistics and their symmetries, in the architecture itself, and these models hold up as machine-learning architectures too. And I measure: a geometry on stimulus space that reads out the information a neural population carries and shows how the code reshapes the structure of the world.

Previously
Research intern with Alex Williams, Flatiron Institute
Research engineer in Brice Bathellier's lab, Institut Pasteur
Summer student with Lorenzo Moneta, CERN
Visiting student with André Longtin and Leonard Maler, University of Ottawa
Co-founder and former president, Machine Learning Journal Club

Research

Three principles of visual coding.

Each starts as a question about the visual system, from the retina to V1 and V4, and turns out to be a question about machine learning too.

Publications

Papers.

Google Scholar ↗

    Community

    Workshops I've co-organized.

    With Sophia Sanborn, Christian Shewmake, Arianna Di Bernardo, Nina Miolane, Bahareh Tolooshams, Chase van de Geijn and Francisco Acosta (NeurReps, SINR), with Samuele Virgili and Gabriel Mahuas (CoSyNe 2024), and with Sonica Saraf and Pietro Zamberlan (CoSyNe 2026).

      News

      Recently.

        Writing

        Early posts.

        Tutorials from 2019–2020, mostly from my Machine Learning Journal Club days.