Adrien Todeschini (@adrtod) header
Adrien Todeschini (@adrtod) profile picture
Bordeaux, France
Joined July 2012
Photo by Adrien Todeschini (@adrtod)
Les journées de #statistique #JDS2024 auront lieu à @Bordeaux du 27 au 31 mai, sur le campus Victoire @univbordeaux ! Évènement d'envergure dans ce domaine de #recherche pour les scientifiques, des événements ouvert à tous sont également proposés ! 👉 inria.fr/fr/journees-de-stat…
1
1
282
📢 Les #JDS2024 c'est lundi ! Au programme, échanges entre scientifiques et événements ouverts au public : ✅️ Soirée débat au Starfish ✅️ Atelier Enseignement de la #statistique ✅️ Atelier Recherche #reproductible✅️ Atelier Stat et #sport inria.fr/fr/journees-de-stat…
3
3
510
Dans le cadre des journées de #statistique #JDS2024, rendez-vous le mardi 28 mai à 19h au bar le #Starfish pour un #Café de la statistique sur la "Prédiction des résultats #électoraux". 📢 Événement gratuit et ouvert au public. 👉 inria.fr/fr/journees-de-stat…
1
2
464
Adrien Todeschini retweeted
MLPs are so foundational, but are there alternatives? MLPs place activation functions on neurons, but can we instead place (learnable) activation functions on weights? Yes, we KAN! We propose Kolmogorov-Arnold Networks (KAN), which are more accurate and interpretable than MLPs.🧵
116
1,019
252
4,990
1,035,130
State Space Models: A Modern Approach This is an interactive textbook on state space models (SSM) using the JAX Python library. probml.github.io/ssm-book/ro…
6
219
4
1,215
207,984
Adrien Todeschini retweeted
Modern LLMs are much more than token predictors. They are also much more than pretrained Transformers. I review the post pretraining processes in my latest blog post: amatriain.net/blog/postpretr…
1
13
2
39
5,184
Adrien Todeschini retweeted
As many of you know, over the past few months I have been sharing Prompt Engineering resources in different forms. I have now compiled them all into a cohesive publication and uploaded to arxiv: arxiv.org/abs/2401.14423
5
91
6
406
41,959
Adrien Todeschini retweeted
I've been avoiding learning CUDA for about 10 years! But it seems like there's an accessible resource out there now. No more excuses 😅!
I used to find writing CUDA code rather terrifying. But then I discovered a couple of tricks that actually make it quite accessible. In this video I introduce CUDA in a way that will be accessible to Python folks, & I even show how to do it all in Colab! youtu.be/nOxKexn3iBo
12
32
487
66,488
Adrien Todeschini retweeted
okay so AI can literally read our minds now. a team from osaka was able to reconstruct visual images from mri scan data using stable diffusion. first row is the image presented to the test subject, second row is the reconstructed image from mri data. wild.
664
6,547
2,416
27,838
7,892,563
Adrien Todeschini retweeted
Wow, my Transformers Catalog has been viewed over 10k since last week's update. Many good comments too. Quick follow up includes link to github for folks who want to file a PR. Also adding a new timeline view where the Y-axis is model size amatriain.net/blog/transform…
10
1
62
7,989
Adrien Todeschini retweeted
Excited to announce our Deep Learning Tuning Playbook, a writeup of tips & tricks we employ when designing DL experiments. We use these techniques to deploy numerous large-scale model improvements and hope formalizing them helps the community do the same! github.com/google-research/t…
26
583
31
2,480
334,433
Adrien Todeschini retweeted
You couldn't make it to #NeurIPS2022 this year? Nothing to worry - I curated a summary for you below focussing on key papers, presentations and workshops in the buzzing space of ML in Biology and Healthcare 👇
22
146
8
754
Adrien Todeschini retweeted
Did you know you can perform ❗exact❗ uncertainty quantification for approximate Gaussian processes in quadratic time? 🤯 Joint work w/ G. Pleiss, M. Pförtner, @PhilippHennig5 and J. Cunningham at #NeurIPS2022. Paper 📄 arxiv.org/abs/2205.15449 Code 💻 github.com/JonathanWenger/it…
2
4
1
38
Adrien Todeschini retweeted
Hi #rstats users! If you’ve been wrangling data in the tidyverse, the Arquero library in #JavaScript will feel familiar. Here's a primer to help translate common wrangling methods from dplyr & tidyr to Arquero, so you can work more nimbly between R & JS 🪄observablehq.com/@observable…
2
43
3
164
Adrien Todeschini retweeted
My AI and ML research paper review time has gone down significantly. Thanks to this amazing tool - explainpaper.com that uses an AI model to explain dense sections In the background, an LLM simplifies and explains complex concepts AI explaining AI 👏👏👏
33
259
28
1,365
Adrien Todeschini retweeted
Our paper on understanding AlphaZero♟️is now published at PNAS! pnas.org/doi/10.1073/pnas.22… The paper "studies" AZ's internals and its behaviors in collaborations with @DeepMind and world chess champion @32gcfhkmm. What did we learn? 🧵
6
127
14
668
Adrien Todeschini retweeted
A Large Language Model trained on scientific papers. Type a text and galactica.ai will generate a paper with relevant references, formulas, and everything. Amazing work by @MetaAI / @paperswithcode
🪐 Introducing Galactica. A large language model for science. Can summarize academic literature, solve math problems, generate Wiki articles, write scientific code, annotate molecules and proteins, and more. Explore and get weights: galactica.org
68
430
90
2,181
Adrien Todeschini retweeted
Thanks for the many comments on my tweet on TabPFNs: teapawt.net/FrankRHutter/status/15… Here's an update on how this may revolutionize data science. TL;DR: learning classification algorithms rather than programming them; real-time AutoML; democratizing ML to tackle *small data*! 🧵1/9
This may revolutionize data science: we introduce TabPFN, a new tabular data classification method that takes 1 second & yields SOTA performance (better than hyperparameter-optimized gradient boosting in 1h). Current limits: up to 1k data points, 100 features, 10 classes. 🧵1/6
3
48
5
236
Adrien Todeschini retweeted
This may revolutionize data science: we introduce TabPFN, a new tabular data classification method that takes 1 second & yields SOTA performance (better than hyperparameter-optimized gradient boosting in 1h). Current limits: up to 1k data points, 100 features, 10 classes. 🧵1/6
102
728
127
4,055
Adrien Todeschini retweeted
Have you ever wanted to add a math description for your #Python function but found it time-consuming to do so? latexify_py allows you to add LaTex math description with only one decorator. Link to latexify_py: github.com/google/latexify_p…
74
1,191
91
6,394