Vincent Lostanlen
@lostanlen.bsky.social
590 followers 160 following 240 posts
Scientist at CNRS. https://audio.ls2n.fr
Posts Media Videos Starter Packs
Pinned
lostanlen.bsky.social
meet the PhD students i'm advising

Matthieu Carreau
Xiran Zhang
Orian Sharoni
Yuexuan Kong
Yasmine Benhamadi
Han Han
Reyhaneh Abbasi (visiting)

picture taken earlier this month in front of Ircam
📸 Mathieu Lagrange

our team website: audio.ls2n.fr
Reposted by Vincent Lostanlen
christhetree.bsky.social
Modulation Discovery with Differentiable Digital Signal Processing

This week I’ll be at @waspaa.com presenting our work on discovering synthesizer modulation signals in arbitrary audio.

arXiv: arxiv.org/abs/2510.06204
web: christhetr.ee/mod_discovery
code: github.com/christhetree/mod_discovery
lostanlen.bsky.social
also the last chapter of "Sounding Human. Music and Machines 1740/2020" by @deirdrelough.bsky.social
lostanlen.bsky.social
hello @ethanhein.bsky.social,

Jim Gabaret "L'art des IA"

interesting, well-balanced and well-sources. there is a chapter on music

out Oct 15
Reposted by Vincent Lostanlen
profannawatts.bsky.social
Lol the Nobels can't even acknowledge women's contribution to discovery. But sure let's acknowledge The Machines.
Headline from an article in Nature this week that states "Prizes must recognize machine contributions to discovery. The future of science will be written by humans and machines together. Awards should reflect that reality."
lostanlen.bsky.social
written with PhD student Yuexuan Kong and her advisors, both at @researchdeezer.bsky.social (G. Meseguer-Brocal, R. Hennequin), and @cnrs.fr (M. Lagrange)

+ special thanks to @ethanhein.bsky.social for this insightful blog post, which we credit in Section II
ethanhein.substack.com/p/identifyin...
Identifying modulations
Warm up the truck
ethanhein.substack.com
lostanlen.bsky.social
We also prove that STONE induces spontaneous symmetry breaking, since modes of limited transposition (e.g., augmented, diminished, whole-tone) are multistable in the circle of fifths, the associated CPSD gradient is driven towards more "tonal" (i.e., asymmetric) scales, e.g., diatonic or pentatonic.
lostanlen.bsky.social
Our main result. If the frequency of cross-power spectral density is set to be coprime with Q=12, STONE does not collapse: the self-supervised objective is well-posed and sparsity-inducing in pitch class space.
lostanlen.bsky.social
"Understanding equivariant self-supervised learning in musical pitch class space" is out on IEEE Signal Processing Letters!

in which we revisit STONE (Kong et al., ISMIR 2024) from a more theoretical standpoint

🔗 hal.science/hal-05297975v1
STONE flowchart
lostanlen.bsky.social
"La mystification du féminin gagne un nouveau degré d'abstraction dans les présupposés du vitalisme, qu'il oublie ou non le féminin."
@azeliefayolle.bsky.social
Reposted by Vincent Lostanlen
deirdrelough.bsky.social
We are in proofs! The Museum of Imaginary Musical Instruments book coming in 2026 @reaktionbooks.bsky.social