Reposted by Max BrodeurFlavio Martinelli @flavioh.bsky.social · 10/06/2026NEW PAPER. Why do larger networks train better? "Because they contain more candidate *sub*networks that can learn the task" → lottery tickets This popular explanation uses an appealing but misleading metaphor🧵 We propose an intuitive alternative grounded in theory: escape dimensions 519953
Reposted by Max BrodeurSimone (she/her) @simonerencken.bsky.social · 31/05/2026I’m happy to share that our paper about the Sepia officinalis genome is out in eLife!🧬 We compared the assemblies of two individuals that had a different predicted number of chromosomes, and found highly repetitive regions that likely prevented a complete assembly. elifesciences.org/articles/107... 02210
Max Brodeur @maxbrodeur.bsky.social · 11/02/2026Strange attractor in 1000D space of the activity in a random recurrent network — viewed from the first 20 principal components (in pairs). #mathart #neuroscience 041
Max Brodeur @maxbrodeur.bsky.social · 04/08/2025Euclid, Newton & Turing This weekend, at the John Rylands Library in Manchester, I finally found a historical copy of Euclid’s Elements. And right next to it: Newton’s Principia and Turing’s handwritten notes. Completely overwhelming. 130
Max Brodeur @maxbrodeur.bsky.social · 31/07/2025Smooth-rolling knots! At @markpauly.bsky.social’s lab, we found a way to optimize knots for smooth rolling. I presented our work at Bridges 2025 – the conference for #MathArt. More below ↓ 2314
Max Brodeur @maxbrodeur.bsky.social · 29/07/2025What is the hidden structure of the natural numbers? In the UMAP paper, @lelandmcinnes.bsky.social et al. embedded the integers as binary vectors of their prime factors. This UMAP visualization of 30 million integers reveals a fractal-like geometry. #UMAP #mathart #dataviz #numbertheory 282