Vincent Lostanlen

Appel à contributions : Assises CPM 2027 Assises de la recherche en cultures populaires et médiatiques, troisième édition – CPM 2027 Cultures POP au(x) présent(s)   Dates: 17 au 19 mars 2027 Lieux: KBR et Université de Namur (Belgique) Échéa...

📣 L'appel à contributions des prochaines Assises de la recherche en cultures populaires et médiatiques est en ligne (17-19 mars 2027, KBR et Université de Namur) ! Les candidatures sont ouvertes jusqu'au 31 mai.
lpcm.hypotheses.org/44160

Torus embeddings Many data representations are vectors of continuous values. In particular, deep learning embeddings are data-driven representations, typically either unconstrained in Euclidean space, or constrained to a hypersphere. These may also be translated into integer representations (quantised) for efficient large-scale use. However, the fundamental (and most efficient) numeric representation in the overwhelming majority of existing computers is integers with overflow -- and vectors of these integers do not correspond to either of these spaces, but instead to the topology of a (hyper)torus. This mismatch can lead to wasted representation capacity. Here we show that common deep learning frameworks can be adapted, quite simply, to create representations with inherent toroidal topology. We investigate two alternative strategies, demonstrating that a normalisation-based strategy leads to training with desirable stability and performance properties, comparable to a standard hyperspherical L2 normalisation. We also demonstrate that a torus embedding maintains desirable quantisation properties. The torus embedding does not outperform hypersphere embeddings in general, but is comparable, and opens the possibility to train deep embeddings which have an extremely simple pathway to efficient `TinyML' embedded implementation.

New from me: "Torus embeddings" - a topological exploration in deep learning and CPUs https://arxiv.org/abs/2603.03135 #preprint #deeplearning #topology

Thinking of Bolaño.

“…all horrors are dulled by routine.”

(By Night in Chile, tr. Chris Andrews)

Bilateral ventral pathways support phonological awareness at reading onset in Spanish-speaking children Abstract. Reading is a fundamental human skill that has been widely studied. While substantial progress has been made in identifying the white matter pathways supporting reading in adults, less is kno...

Phonological processing = left dorsal stream?
Not in Mexican Spanish-speaking children at reading onset.
Here, bilateral ventral — not dorsal— pathways predict phonological awareness.
Neuroscience should broaden the populations it studies to build generalizable brain models.
doi.org/10.1162/NOL....

**NEW POST**

LLMs hallucinate, & they hallucinate WORSE on topics we don't already know/write a lot about. This creates a huge accuracy gap.

This week I tested ChatGPT on Beethoven & Smyth to show how big this gap is & why it matters. It made up a LOT.

leahbroad.substack.com/p/chatgpts-a...

Delighted to share our new lab page from the Music Cognition Lab! We have lot's of events and research projects coming up, so you can stay updated here!

#MusicPsychology

Fantastic post by Colin Raffel, "We Are Over-Indexing on Paper Acceptance," drafted in May 2021 (!) but only posted now. The more things change..

Last sentence: "If you want to judge a researcher’s quality, the only meaningful way is to read their papers and judge for yourself."

"you've got to dig it to *dig* it, you dig?"

16 composing women you couldn't live without

1) Radigue
2) Ustvolskaya
3) Bokanowski
4) Thorvaldsdottir
5) M. Monk
6) Crawford-Seeger
7) Hildegard van Bingen
8) Tailleferre
9) Borzelli
10) Ann Southam
11) Maconchy
12) Helena Gough
13) Fanny Mendelssohn
14) Saariaho
15) Falcinelli
16) Mel Bonis

Jan Jelinek
Loop Finding Jazz Records
2026 Pressing

boomkat.com/products/loo...

Linkedin post by Downtown Music that says "It's official. Virgin Music Group and Downtown Music have joined forces, united in our commitment to the global independent community."

Love when two entities owned by Universal Music Group join forces to support the global independent music community.

Rosmarie Waldrop:

“In crossing the Atlantic my phonemes settled somewhere between German and English. I speak either language with an accent. This has saved me the illusion of being the master of language. I enter it at a skewed angle, through the fissures, the slight difference.”

SCRAPL: Scattering Transform with Random Paths for Machine Learning

I’m excited to share our #ICLR2026 paper on SCRAPL: an algorithm that makes wavelet scattering transforms usable as differentiable loss functions!

paper: openreview.net/forum?id=RuYwbd5xYa
web: christhetr.ee/scrapl/

Next time you want to use a vibe coded compile think about benchmarking the resulting binary. ccc on SQLite is up to 150k slower on some SQL queries and the benchmark took 2 hours vs 10sec for GCC compiled harshanu.space/en/tech/ccc-...

Hammer, meet nail: Suno is "a vending machine, not a kitchen."

MANIFESTE26: AUDITEURS LIBRES
Profitez de tous les privilèges qu'offre le statut d'auditeur : suivre les cours des artistes invité·e·s, accéder à l’exceptionnel fonds documentaire de la médiathèque de l'Ircam, se rendre gratuitement à toutes les sorties d'ateliers et master classes de l'académie.

What a Rare Condition Can Teach Us About the Power of Music The study of musical anhedonia—the inability to enjoy music—is revealing how music moves us.

Thrilled to share that The New Yorker featured my work on musical anhedonia - why some people don’t experience pleasure from music, and what this reveals about how our brains predict and process reward.
www.newyorker.com/culture/anna...
#neuroskyence #musicskyence #psychscisky

Je suis l'auteur.

C'est un rappel que nous cotisons tous pour offrir des rentes illimitées aux GAFAM qui peuvent ainsi financer leur avance sur de nouveaux enjeux technos de pointe et empêcher une concurrence d'émerger, jusqu'à siphonner nos propres compétences. Résoudre ça est vital pour notre

Hourly Comics Day 2026 Part 1 of 2

Sometime the day before My two cats are communing with my baby. The white cat (Lady) says "Ok, this one is for waking them up first. Rock, paper, scissors–". In the second frame we see them each put a paw in and the baby has made scissors. Lady says "Ah, you win again".

12:30am I am softly humming Chopin to my baby. She is wide awake "eeeeeee", acquiring consciousness

7:30am I look after Plover (not her real name) most mornings so Mélodie can get more sleep. The baby is pictured saying "ee-ayagoo ayyyyyyyy" on the bed. We have our little rituals. I stand with her in front of the window saying "good morning, spruce!" Recently I've rediscovered my love for my mocha pot. It's a quiet way to make coffee, one-handed.

Hourly Comics Day 2026 comics dump!

The Birdsong Lab is recruiting graduate students! Links in the comments. Please repost!

Bridging Brains and Bioacoustics Seminar announcement for Friday, January 30th at 12:30 ET From Audrey Looby, University of Victoria: "The Prevalence and Important of Soniferous Fishes" From Brooke Vetter, University of St. Thomas "Anthropogenic Noise and Fish Hearing: From Ecological Impacts to Invasive Species Management"

📣 New Bridging Brains and Bioacoustics Seminar next week!

Join us to learn about the ecology and neuroscience of soniferous fish from Audrey Looby and Brooke Vetter
🐟🐠🐡

@FishSounds.bsky.social

#bioacoustics
#neuroskyence
#prattle 💬

🗓️ January 30th, 12:30 ET
✅ Register here: www.braincoustics.com

> Now in 2026, it is actually getting to be somewhat more common for jazz musicians, especially younger, more experimental ones, to [use] the recording studio or the digital audio workstation as the creative canvas complementing live interplay and improvisation in the room.