"Make sure that your model can overfit on small training set" might be the single best sanity check when building ML models. It helped me solve countless implementation errors and to better understand capacity. I first heart it from @fchollet, thanks!
Everything you wanted to know about activation functions, but were afraid to ask!
Turns out there has been at least 4⃣ 0⃣0⃣ activation functions proposed over 3⃣decades and there is a paper that reviews all of them!
https://t.co/PapXkkr2hI
With the rise of AI-generated "fake" human content--"deepfake" imagery, voice cloning scams & chatbot babble plagiarism--those of us working on social impact @huggingface put together a collection of some of the state-of-the-art technology that can help:
https://t.co/nFS7GW8dtk
let me click that tweet to see what other people are saying about it
unrelated meme
unrelated meme
ad
unrelated meme
verified rightwing troll
ad
unrelated meme
[#Résultat travaux de recherche]
👀 Imaginez une voix capable de restituer un texte automatiquement en #breton
👌avec la bonne prononciation & intonation
voire capable de parler directement en s'adaptant au contexte ?
⏬ lire la suite ...😀
BTW @NeurIPSConf you should put the keynote recordings available for everybody.
I really doubt it would hurt revenue in any way, that would be far more in line with what the research community is [supposed to be] about, and that'd do a lot for equity egality etc.
This is a reminder that you probably spent money on chess courses that you have not opened in months, instead choosing to play blitz until you lose 200 rating points like a degenerate
L'hôpital public s'effondre.
On a besoin de soigner des gens et vous nous envoyez un comptable.
➡️Augmentation du prix des consultations
➡️Pénurie de médicaments et augmentation des prix
➡️ Effondrement de l'hôpital
Tout le monde est contre vous.
Donc ENCORE un nouveau 49.3.
Which one is the better machine learning interpretation method?
LIME or SHAP?
Despite the gold rush in interpretability research, both methods are still OGs when it comes to explaining predictions.
Let's compare the giants.
NLP research over the years:
1990s: specialized features for NER
2000s: non-parametric prior for unsupervised parsing
2010s: embeddings + a forward/backward, inside/outside RNN
Today: When you say “Super pretty please” to the LLM, performance goes up
This is not sad, this is a warning. A company did everything (mostly) right: created a vitally useful free service, worked hard to keep it healthy, shared data for research. Then another company swooped the data and made it compete against itself.
Rappel important : le maximum de la pluie d’étoiles filantes Perseides, l’une des plus belles de l’année, sera le 12,13 et 14 août avec 70 a 110 étoiles filantes par heure environ ! 💫💫
Le gouvernement a choisi de dissoudre @lessoulevements qui représentent aujourd'hui + de 110 000 personnes. Cette décision alarmante est le fruit d'une escalade répressive envers les militants écologistes qui s'accentue depuis 2015 et inquiète les experts de l'ONU. Un 🧶 1/14