Note
TabICL is a state-of-the-art tabular learner [Qu et al 2025]. The key is its very rich prior, that is baked in a pre-trained architecture -a table foundation model-, and leveraged by in-context-learning. Thanks to clever choices, it is fast and scalable, efficient even without a GPU.
Contents
Modern AIs acquire new capabilities by combining tools to perform a complex task, controlling them like an agent. Unlike traditional programming, they define the sequences of actions themselves.
Note
This post was originally published in French as part of my scientific chronicle in Les Echos.
Modern AIs are increasingly using …
The key to the most powerful conversational AIs is to reason by breaking down a complex task into simpler subproblems. Why is this crucial, and how does it work?
Note
This post was originally published in French as part of my scientific chronicle in Les Echos.
The recent release of …
I would like to take a brief moment to reflect on what drives me as an academic.
Academia’s root are in creating knowledge and sharing it. We, academics, have a role to play in shaping society. In computer science, we sometimes focus on the creation of technology. Here, creation …
Since 2017, an AI has been defeating the best Go experts, despite the game being particularly challenging. Such “super intelligence” is rare, but it could also emerge in fundamental mathematics.
Note
This post was originally published in French as part of my scientific chronicle in Les Echos.
Is unbiased data important to build health AI? Yes!
Can there be unbiased data? No!
The notion of bias depends on the intended use.
In medicine, we have seen the importance of tuning devices and decisions for the target population. The problem is not …
Note
For me, 2024 was full of back and forth between research, software, and connecting these to society. Here, I lay out some highlights on AI and society, as well as research and software, around tabular AI and language models.
As 2025 starts, I’m looking back on 2024. It …
Improving conversational artificial intelligences or simpler prediction engines involves overcoming biases, that is, going beyond the limits of data. But the notion of bias is subtle, as it depends on the goals.
Note
This post was originally published in French as part of my scientific chronicle in Les Echos.
In …
Despite their apparent intelligence, conversational artificial intelligences often lack logic. The debate rages on: do they reason or do they recite snatches of text memorized on the Internet?
Note
This post was originally published in French as part of my scientific chronicle in Les Echos. I updated it with new …
Note
Foundation models, pretrained and readily usable for many downstream tasks, have changed the way we process text, images, and sound. Can we achieve similar breakthroughs for tables? Here I explain why with “CARTE”, we’ve made significant headway.
Contents
We just released skrub 0.2.0. This release markedly simplifies learning on complex dataframes.
The highlight of the release is the tabular_learner function, which facilitates creating pipelines that readily perform machine learning on dataframes, adding preprocessing to a scikit-learn compatible learner …
Note
Open-source efforts around scikit-learn at Inria are spinning off to a new enterprise, Probabl, in charge of sustainable development of a data-science commons.
Contents
Note
François Chollet rightfully said that people often underestimate the impact of scikit-learn. I give here a few illustrations to back his claim.
A few days ago, François Chollet (the creator of Keras, the library that that democratized deep learning) posted:
Indeed, scikit-learn continues to be the most popular machine …
English summary
I have been appointed to the government-level panel of experts on AI, to set the national vision and strategy in France.
J’ai l’honneur d’être nommé au comité de l’intelligence artificielle du gouvernement Français.
La mission qui nous est confiée d’éclairer l’action publique …
A retrospective on last year (2022): I embarked on a new scientific adventure, assembling a team focused on developing machine learning for health and social science. The team has existed for almost a year, and the vision is nice shaping up. Let me share with you illustrations of where we …
The Mayavi Python software, and my personal history: A thread on Python and scipy ecosystems, building open source codebase, and meeting really cool and friendly people
I am writing today as a goodbye to the project: I used to be one of the core contributors and maintainers but have been …
Broad decoding models that can specialize to discriminate closely-related mental process with limited data
TL;DR
Decoding models can help isolating which mental processes are implied by the activation of given brain structures. But to support a broad conclusion, they must be trained on many studies, a difficult problem given …
Note
Join us to work on reinventing data-science practices and tools to produce robust analysis with less data curation.
It is well known that data cleaning and preparation are a heavy burden to the data scientist.
In the dirty data project, we have been conducting machine-learning research …
Note
With the growth of scikit-learn and the wider PyData ecosystem, we want to recruit in the Inria scikit-learn team for a new role. Departing from our usual focus on excellence in algorithms, statistics, or code, we want to add to the team someone with some technical understanding, but an …
The year 2020 has undoubtedly been interesting: the covid19 pandemic stroke while I was on a work sabbatical in Montréal, at the MNI and the MILA, and it pushed further my interest in machine learning for health-care. My highlights this year revolve around basic and applied data-science for health.
Highlights …
