Deep generative models parametrize very flexible families of distributions able to fit complicated datasets of images or text. These models provide independent samples from complex high-distributions at negligible costs. On the other hand, sampling exactly a target distribution, such the Boltzmann distribution of a physical system, is typically challenging: either because of dimensionality,...
Quantum computing aims to leverage the principles of quantum mechanics, such as superposition, to encode and process information in ways that classical computers cannot, potentially handling exponentially larger amounts of information. However, harnessing this computational advantage requires quantum algorithms capable of encoding data into superpositions and providing answers with minimal...
Measuring the uncertainty associated to a model's prediction is a central part of statistical practice. In the context of modern deep learning practice, several methods for quantifying the uncertainty of neural networks co-exist. Yet, theoretical guarantees for these methods are scarce in the theoretical literature. In this talk, I will discuss how some of them compare in a mathematically...
Riboswitches are structured allosteric RNA molecules that change conformation in response to a metabolite binding event, eventually triggering a regulatory response. Computational modelling of the structure of these molecules is complicated by a complex network of tertiary contacts, stabilized by the presence of their cognate metabolite. In this work, we focus on the aptamer domain of SAM-I...
Physicists routinely need probabilistic models for a number of tasks such as parameter inference or the generation of new realizations of a field. Establishing such models for highly non-Gaussian fields is a challenge, especially when the number of samples is limited. In this paper, we introduce scattering spectra models for stationary fields and we show that they provide accurate and robust...
Climate models and Numerical Weather Prediction (NWP) Models describe the atmospheric circulation with a
limited resolution. There unavoidably remains processes that involve spatial scales shorter than the
grid scales, ie processes that are unresolved. Cloud processes, turbulence near the surface and internal
gravity waves propagating from lower to upper layers are among the main dynamical...
In this talk I will first introduce the basics of radar imaging and present some applications
for climate science. I will then show how machine learning can make a key contribution
to improve radar data degraded by the speckle phenomenon and extract useful information.
I will focus on self-supervised methods allowing for exploiting a wide range of unlabeled data.
Data assimilation is a central problem in many geophysical applications, such as weather forecasting. It aims to estimate the state of a potentially large system, such as the atmosphere, from sparse observations, supplemented by prior physical knowledge. The size of the systems involved and the complexity of the underlying physical equations make it a challenging task from a computational...
Large astronomical facilities generate an ever-increasing data volume, rapidly approaching the exascale, following the need for better resolution, better sensitivity, and larger wavelength coverage. Modern radio astronomy is strongly affected, especially regarding giant radio interferometers that produce large quantities of raw data. In particular, the forthcoming arrival of the SKA (Square...
The field of experimental astronomy is entering an exciting new era, with the emergence of extremely large telescopes, hosts to primary mirrors the size of several basketball courts. Among the many challenges associated with the construction and operations of such giant scientific infrastructures, the complexity of embedded computing facilities is notably heavy. In particular, the real-time...
Recent years have witnessed intense interactions between cognitive neuroscience and artificial intelligence, with the deep learning revolution driving new developments in neuroscience.
A first aspect concerns the processing of neuroscience data, which is often in the form of time courses. These data are often short and noisy, and suffer from poorly controlled confounding effects. AI-powered...
Do representations proposed in linguistic theories, such as constituent trees, correspond to actual data structures constructed in real-time in the brain during language comprehension? And if so, what are the brain regions involved? This question was investigated in a series of functional magnetic resonance studies using various experimental paradigms, including repetition priming, syntactic...
A machine readable and verifiable account of a large portion of human mathematics would change the way mathematicians can work, learn and collaborate. While impressive progress has been made in the mathematical standard libraries of proof assistants like Lean, Isabelle and Coq, the proportion of mathematical results formalized in such systems remains tiny overall. In the talk, I will argue...
Single-cell data constitute a major breakthrough in life sciences. Their integration will enable us to investigate outstanding biological and medical questions thus far inaccessible. However, still few methods exist to integrate different single-cell modalities, corresponding to omics data (e.g. DNA methylation, proteome, chromatin accessibility), plus spatial positioning and images....
The combination of artificial intelligence and the increasing digitization of the health sector opens up perspectives for using data for research and daily decision-making tools for patients and healthcare providers. However, the systematic deployment of these technologies requires better control of their performance, particularly in terms of generalization and explainability. These notions...
In this work we adapt recent model reduction approaches to predict the solutions of
time-dependent parametrized problems describing crowd motion in the presence of ob-
stacles. The problem of interest is a discrete contact model, which is formulated as a
constrained least-squares optimization statement. The parametric variations in the prob-
lem (associated with the geometric configuration...
Generative models have started to integrate into the scientific computing toolkit. One notable instance of this integration is the utilization of normalizing flows (NF) in the development of sampling and variational inference algorithms. This work introduces a novel algorithm, GflowMC, which relies on a Metropolis-within-Gibbs framework within the latent space of NFs. This approach addresses...
