I'm 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 in Paris.
Website · Google Scholar · X · Bluesky · LinkedIn · simone . azeglio at gmail . com
My work has two sides: building the structure of natural scenes into models of visual neurons, and measuring what information the resulting codes carry.
| Question | Work | |
|---|---|---|
| Statistics | What do neighbouring pixels say together? | Higher-order convolution (NeurIPS 2025) · and in the retina |
| Symmetry | How should a code change when the world does? | Scale-equivariant retinal codes support prey hunting (CoSyNe 2026) |
| Information | Which bits, about what? | Multi-scale information geometry (NeurIPS 2026 Oral) · Stimulus-specific information (NeurIPS 2025 Spotlight) · Score-based Riemannian metric |
- Sep 2026 · NeurReps 2024 and 2025 proceedings published as PMLR volume 282, which I co-edited
- Jul 2026 · Defended my PhD thesis at the Vision Institute (Sorbonne) and ENS Paris
- Mar 2026 · Co-organizing a workshop on modern efficient coding at CoSyNe 2026
- Mar 2026 · CoSyNe poster 1-146: mouse OFF-α ganglion cells are scale equivariant
- Dec 2025 · NeurReps 2025, our 4th edition, is a wrap
Synced daily from my website.
Co-organizer of NeurReps at NeurIPS (2022–2025), ECMA at CoSyNe 2026, Sharpening Our Sight at CoSyNe 2024 and SINR at Bernstein (2022–2023). Co-founder of the Machine Learning Journal Club. Before the PhD: Flatiron Institute, Institut Pasteur, CERN and the University of Ottawa.
Always glad to talk about collaborations, and to mentor students whose interests line up with these topics.
The banner is generated by scripts/make_banner.py: OFF and ON ganglion-cell mosaics with difference-of-Gaussians receptive fields, the same model as the header of my website. The activity card is refreshed daily by scripts/make_stats.py.