26ème Congrès Général de la SFP

Session

Mini-colloques: MC10 Physique et intelligence artificielle

3 juil. 2023, 16:45

Centre des Congrès de la Villette (Cité des sciences et de l'Industrie, Paris)

27. Machine Learning prediction of the structural/textural-mechanical properties relationship of pyrolytic carbons

Franck Polewczyk (CEA DAM-DIF & Institut des Sciences Moléculaires (ISM))

03/07/2023 16:45

MC10 Physique et intelligence artificielle

Contribution orale

The ability of C/C composites to maintain high mechanical properties up to very high temperatures (above 3000 K), combined with their low densities, justifies their use in extreme conditions, especially in aerospace and defense applications. However, due to the very large anisotropy of their constituents, their partial crystalline order, and the difficulty to perform experiments under the...

162. Chemically-specific multiscale mechanical simulation of nanomaterials: improving tractability with machine learning

Maxime VASSAUX (Institut de Physique de Rennes, Univ. Rennes, CNRS)

03/07/2023 17:05

MC10 Physique et intelligence artificielle

Contribution orale

Material properties, particularly mechanical ones, are intrinsically multiscale, emerging from the material's structure ranging from the atomic to the continuum structure. Multiscale modelling and simulation based on the assumption of a separation of scales in the architectures of nanomaterials appears as a promising alternative to trial-and-error experiments for material design. The current...

455. agnostic machine learning description of chemical reactions in solution

Timothée Devergne (IMPMC - Sorbonne université)

03/07/2023 17:25

MC10 Physique et intelligence artificielle

Contribution orale

We address the problem of the training of a machine learning potential (MLP) for the atomistic study of rare events. By training a MLP only on a few dozens of agnostic out of equilibrium ab initio trajectories shot from the top of the barrier, we are able to recover free energies close to ab initio accuracy. As well as showing that it is possible to recover transition rates and...

359. Physics-Informed Neural Networks for Learning about Magnets and their dynamics

M. Matthieu Carreau (Télécom Paris, Institut Polytechnique de Paris)

03/07/2023 17:45

MC10 Physique et intelligence artificielle

Contribution orale

With the rapid development of miniaturized devices in spintronics, the dynamics of nanomagnets is of both theoretical and practical interest. The equations of motion for a magnetic moment that figure out the average magnetization embedded in a medium, are differential equations, but contain time derivatives on both sides, that cannot be-in general-recast in a form that is useful for usual...

317. Méthodes d’apprentissage machine dans l’expérience JUNO de physique des neutrinos.

Benoit VIAUD (CNRS - in2p3 - Subatech)

03/07/2023 18:05

MC10 Physique et intelligence artificielle

Contribution orale

L’expérience JUNO (Jiangmen Underground Neutrino Observatory), en cours de construction en Chine, sera l’un des plus grands détecteurs de neutrinos de prochaine génération. Si son champ d'investigation est multiple, l’un des ses principaux buts concerne la hiérarchie de masse des neutrinos, inconnue à ce jour. Sa détermination s’appuie grandement sur la mesure du spectre en énergie...

416. Machine Learning for Real-Time Processing of ATLAS Liquid Argon Calorimeter Signals with FPGAs

Dr Nairit SUR (CPPM - CNRS/IN2P3)

03/07/2023 18:25

MC10 Physique et intelligence artificielle

Contribution orale

The Phase-II upgrade of the LHC will increase its instantaneous luminosity by a factor of 7 leading to the High Luminosity LHC (HL-LHC). At the HL-LHC, the number of proton-proton collisions in one bunch crossing (called pileup) increases significantly, putting more stringent requirements on the LHC detectors electronics and real-time data processing capabilities.

The ATLAS Liquid Argon...

420. Accélération des simulations d’oscillateurs spintroniques à l’aide d’une approche non-conventionnelle guidée par les données

Simon de Wergifosse (UCLouvain)

06/07/2023 08:30

MC10 Physique et intelligence artificielle

Contribution orale

Le calcul neuromorphique est un champ récent de l’intelligence artificielle visant à diminuer la consommation énergétique des tâches cognitives résolues en s’inspirant de l’architecture du cerveau humain. Dans ce contexte, les nano-oscillateurs à transfert de spin ont récemment fait l’objet d’une grande attention, notamment en raison de leur utilisation potentielle en tant que neurones...

258. Efficient estimation of trainability for variational quantum circuits

Valentin Heyraud (Université Paris Cité, CNRS, Matériaux et Phénomènes Quantiques (MPQ))

06/07/2023 08:50

MC10 Physique et intelligence artificielle

Contribution orale

Variational quantum algorithms are emerging as promising tools to tackle complex problems ranging from quantum chemistry to combinatorial optimization. These algorithms optimize the continuous parameters of a variational quantum circuit as to extremize a cost function encoding the task at hand. The optimization is achieved through gradient descent methods. Such models can suffer from the...

498. ChatGPT, a brilliant but messy intern or: how can ChatGPT be used seriously in a physicist's daily work?

M. David Rousseau (IJCLab, Orsay, France)

06/07/2023 09:10

MC10 Physique et intelligence artificielle

Contribution orale

ChatGPT, a Large Language Model based chatbot with astonishing capabilities, has become very popular on social networks in particular. But beyond amazing success and dismal failure anecdotes, to what extent can ChatGPT (or Bing) be used realistically in the physicist's daily life?
The experiments of twenty particle physicists have been collected as user stories. They are summarised in this...

491. Trajectographie par apprentissage géométrique profond pour faire face au déluge de données attendu au CERN

Sylvain CAILLOU (L2I Toulouse, CNRS/IN2P3, UT3)

06/07/2023 09:30

MC10 Physique et intelligence artificielle

Contribution orale

Le boson de Higgs a été découvert auprès du collisionneur LHC au CERN en 2012. Afin de mieux comprendre les propriétés du boson de Higgs et l’origine ultime de la matière les chercheurs ont maintenant besoin de beaucoup plus de données. La phase dite de haute luminosité (HiLumi-LHC) va démarrer en 2029 et doit permettre de multiplier par 10 le taux de production de bosons de Higgs afin...

259. Deep learning for sparse spectral ptychographic x-ray computed tomography

M. Redhouane BOUDJEHEM (Univ. Grenoble Alpes, CNRS, Grenoble INP, Institut Néel, Grenoble, France)

06/07/2023 09:50

MC10 Physique et intelligence artificielle

Contribution orale

X-ray nanotomography is a well-established technique with many applications in material science and biology. The spatial resolution of classical CT can be enhanced when using ptychographic projections [1] measured at different angles to reconstruct a 3D volume.
X-ray Ptychography [2][3] is a coherent diffraction imaging technique based on the acquisition of multiple diffraction patterns...

501. Reconstruction of electromagnetic showers in calorimeters using Deep Learning

Polina Simkina

06/07/2023 10:10

MC10 Physique et intelligence artificielle

Contribution orale

Machine Learning (ML) algorithms are currently a leading choice for Data Analysis applications in various fields: from industry to science and medicine. Following the general trend, different ML methods (Boosted Decision Trees, Neural Networks) have already been successfully used for data reconstruction and analysis in the CMS experiment. More sophisticated algorithms are becoming available,...