Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : Sorbonne Universite – Institut des Systèmes Intell
Durée : 36 mois
Contact : patrick.gallinari@sorbonne-universite.fr
Date limite de publication : 2023-05-26

Contexte :
Physics-aware deep learning is an emerging research field aiming at investigating the potential of AI methods to advance scientific research for the modeling of complex natural phenomena. This research topic investigates how to leverage prior knowledge of first principles (physics) together with the ability of machine learning at extracting information from data. This is a fast-growing field with the potential to boost scientific progress and to change the way we develop research in a whole range of scientific domains. An area where this idea raises high hopes is the modeling of complex dynamics characterizing natural phenomena occurring in domains as diverse as climate science, earth science, biology, fluid dynamics, etc.

Sujet :
The objective of the PhD project is the development of Physics-aware deep learning methods for the modeling of complex spatio-temporal dynamics. The direct application of state-of-the-art deep learning (DL) methods for modeling and solving physical dynamics occurring in nature is limited by the complexity of the underlying phenomena, the need for large amounts of data and their inability to learn physically consistent laws. This has motivated the recent exploration of physics-aware methods incorporating prior physical knowledge. Although promising and rapidly developing, this research field faces several challenges. For this PhD project we will address two main challenges, namely the construction of hybrid models for integrating physics with DL and generalization issues which condition the usability of DL for physics.

— Integrating DL and physics for spatio-temporal dynamics forecasting and solving PDEs

In physics and many related fields, partial differential equations (PDEs) are the main tool for modeling and characterizing the dynamics underlying complex phenomena. Combining PDE models with ML is a natural idea when building physics-aware DL models and it is one of the key challenges in the field. This has been explored for two main directions: (i) augmenting low resolution solvers with ML in order to reach the accuracy of high-fidelity models at a reduced computational cost, and (ii) complementing incomplete physical models with ML by integrating observation data through machine learning. A first direction of the PhD will then be to investigate hybrid physics-DL models using the recently proposed framework of neural operators. The latter opens the possibility of combining and learning multiple spatio-temporal scales within a unified formalism, a challenge in DL.

— Domain generalization for deep learning based dynamical models

Explicit physical models come with guarantees and can be used in any context (also called domain or environment) where the model is valid. These models reflect explicit causality relations between the different variables involved in the model. This is not the case for DL: statistical models learn correlations from sample observations, their validity is usually limited to the context of the training domain. This is a critical issue for the adoption of ML for modeling the physical world. In relation with the construction of hybrid models as described above, one will investigate this issue along two main directions. The first one is a purely data-based approach and exploits ideas from learning from multiple environments through task decomposition. The second one, takes a dual perspective, relying on prior physical knowledge of the system equations and directly targets the problem of solving parametric PDEs, exploiting ideas from meta-learning.

Profil du candidat :
Computer science or applied mathematics. Good programming skills.

Formation et compétences requises :
Master degree in computer science or applied mathematics, Engineering school. Background and experience in machine learning.

Adresse d’emploi :
Sorbonne Université (S.U.), Pierre et Marie Campus in the center of Paris. The candidate will integrate the MLIA team (Machine Learning and Deep Learning for Information Access) at ISIR (Institut des Systèmes Intelligents et de Robotique).

Document attaché : 202304251022_2023-04-PhD-Description-Physics-Aware-Deep-Learning.pdf

Offre en lien avec l’Action/le Réseau : DOING/– — –

Laboratoire/Entreprise : MaIAGE/Wimmics
Durée : 36 months
Contact : claire.nedellec@inrae.fr
Date limite de publication : 2023-06-30

Contexte :
We are seeking a highly motivated PhD candidate within the framework of the research project “Information acquisition from textual data for early insect vector surveillance inplant health”. The central aim of the thesis project is to develop ontology-based NLP methods to acquire updated and relevant knowledge on epidemic diseases for improved risk management linked to insect vectors. The quality and relevance of the extracted information will be both derived from the collected documents and from the formal representation of the knowledge base. For plant health, the biological interaction between insect vectors, pathogens, and host plants is the primary focus.
The studentship will be affiliated with the laboratory MaIAGE [2] at the INRAE research center in Jouy-en-Josas University of Paris-Saclay (Computer Science GS), with PHIM [3] at the INRAE research center in Montpellier, and Wimmics group, I3S at Univ Côte d’Azur and Inria [4].

[1] https://maiage.inrae.fr/fr/bibliome
[2] https://umr-phim.cirad.fr/en/recherche/comprendre-les-epidemies-dans-les-champs-prism/equipe-forisk
[3] https://team.inria.fr/wimmics/

Sujet :
For more details, see https://maiage.inrae.fr/fr/node/2726

Profil du candidat :
A successful candidate will have an MSc or equivalent in Artificial Intelligence. Knowledge of natural language processing and Knowledge representation methods will be an advantage. Master studies in a related area, e.g., biology or bio-informatics will be an advantage.
– High level of academic English or French, both written and spoken;
– Good programming skills in Python or Java (and preferably experience with deep learning tools)
– Capacity to work as part of a team in a multidisciplinary framework.
– Experiences of applied research to Life Science is an asset.
We offer a motivating research environment with many opportunities for in-house, national and international collaborations and access to computing resources and state-of-the-art research equipment.

Formation et compétences requises :
Application
——
Interested candidates should send their application file to Claire Nédellec (claire.nedellec@inrae.fr), to Nicolas Sauvion (Nicolas.sauvion@inrae.fr), Catherine Faron (faron@i3s.unice.fr).
Applications will be assessed as they are received and decisions taken on a rolling basis. The application should comprise:
– a CV (max 5 pages) with transcripts (Master), diplomas, internships
– a cover letter
– the names and contact of two referees for reference letters< [1] https://maiage.inrae.fr/fr/bibliome [2] https://umr-phim.cirad.fr/en/recherche/comprendre-les-epidemies-dans-les-champs-prism/equipe-forisk [3] https://team.inria.fr/wimmics/

Adresse d’emploi :
Location:

The student will share his/her time between MaIAGE (Jouy-en-Josas) and Wimmics (Sophia Antipolis) over long periods (1 to 2 years) and will make regular visits to PHIM (Montpellier).

Salary:

2300 gross salary per month including social security package (healthcare, pensions, unemployment benefits).

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : Laboratoire LIS, Université de Toulon
Durée : 36 mois
Contact : adeline.paiement@univ-tln.fr
Date limite de publication : 2023-05-31

Contexte :
Two laboratories are involved in the supervision of this PhD: the LIS (Laboratoire d’Informatique et Systèmes) and the Strasbourg Astronomical Observatory. The PhD student will be based a t LIS in Toulon and will work in close collaboration with the Strasbourg Astronomical Observatory.

Applications are to be sent via the CNRS portal: https://emploi.cnrs.fr/Offres/Doctorant/UMR7020-ADEPAI-001/Default.aspx

Sujet :
Context
Observations and image analysis are essential for astrophysics research. Combined with numerical simulations, they are the only means to study the physics that governs astronomical objects. The visual characterisation of Galaxy Morphologies (GM) illustrates this paradigm, as they are correlated with galaxies’ physical parameters and may provide insights into the physical processes that drove their evolutionary history. In particular, the diffuse tidal stellar features that surround galaxies result from past galaxy collisions and bring a testimony on the past mass assembly of galaxies. However, recent or near-future observation missions (e.g. DES, UNIONS, Euclid, LSST, Roman, ARRAKIHS) have collected or are about to collect a rapidly increasing amount of data that can no longer be analysed manually, even via crowd sourcing. This situation can only be resolved by developing automatic image analysis techniques.

Aims & method
The goal of this project is to produce automatic visual analysis tools for finely characterising GMs. In particular, we will focus on the faint tidal debris that have been largely overlooked by previous studies, but that are essential in understanding the collisional history of galaxies. Contrary to previous studies that simplified GM characterisation through discretising GM parameters, we will perform a regression of continuous parameters of interest. This will be achieved through a segmentation of the tidal structures. Previous studies used off-the-shelf deep neural networks (DNN) which were limited in the achievable accuracy for detecting the faint and diffuse tidal structures. To overcome the challenges associated with localising
them, we will develop purpose-designed DNNs that integrate knowledge on the relationships between structures. This knowledge will constrain the learning and increase the localisation robustness. We will endeavour to making the methods applicable on a large scale to major observational missions such as CFIS, Euclid and LSST. Different missions operate at various (numbers of) imaging bands. This raises strong challenges to the practical applicability of automated visual analysis to all available imaging bands. We will explore transfer learning options to adapt the developed methods to the data from various instruments. This project will build on a DNN that was previously designed in our lab, and that is particularly sensitive to the texture and orientation of filamentary structures (Richards et al., 2022b). During the first year of the PhD, we will adapt this DNN to the detection and the segmentation of galactic structures, with a focus on low surface brightness tidal features. During the second year, we will further adapt the DNN’s architecture in order to account for known relationships between galactic structures. The third year will be devoted to preparing the application of the developed methods to real observational data, with the aim to release a tool for the astrophysics community.

Profil du candidat :
Ability to work across disciplines, communication, flexibility and adaptability, analytical and critical thinking, autonomy, teamwork

Formation et compétences requises :
Master level degree in data science, AI, or closely related subject, or in astrophysics with strong experience in programming. Proficient with python and scientific python libraries. An experience with deep learning and/or image processing is recommended.

Adresse d’emploi :
Laboratoire LIS, Université de Toulon, Campus de La Garde, avenue de l’Université, 83130 La Garde

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : LITIS / INSA Rouen Normandie
Durée : 36 mois
Contact : alexandre.pauchet@insa-rouen.fr
Date limite de publication : 2023-05-05

Contexte :
La croissance du nombre d’objets connectés, des robots assistants et des interfaces humain-machines intuitives et naturelles a permis une démocratisation croissante des systèmes cyber-physiques et socio-techniques. Il s’agit de systèmes intégrant à la fois des utilisateurs
humains, des robots et des agents artificiels, en interactions sociales dans desenvironnements réels, augmentés ou virtuels.
Si les interactions avec un seul utilisateur en tour par tour disposent d’une littérature abondante, tous les verrous scientifiques et techniques ne sont pas encore levés concernant les interactions dialogiques. La mise en place de systèmes coopératifs intégrant plusieurs utilisateurs humains, agents virtuels et robots reste difficile. Depuis l’avènement de l’informatique affective (Picard, 1997), des Agents Conversationnels Animés -ACA- (Cassel,
2000), et des robots sociaux (Gockley et al., 2005), qui mettent en œuvre des systèmes prenant en compte les émotions des utilisateurs et pouvant en « jouer », n’est que très peu traitée dans un contexte multi-partie. En particulier, la capacité d’un agent collaboratif à répondre « just in time » à son contexte est peu développée : la plupart des systèmes attendent la fin de chaque tour de parole avant d’interpréter les énoncés des utilisateurs et de décider de la prochaine action communicative à effectuer. Ils ont ainsi un temps de réaction bien supérieur à celui d’un humain, introduisant des ruptures défavorables à l’interaction.

Sujet :
L’objectif de cette thèse est d’étudier les moyens théoriques et pratiques pour permettre une interaction multimodale et multi-parties en temps quasi réel afin d’en améliorer la fluidité et l’acceptabilité. Ces travaux comportent deux originalités principales : (1) la gestion au fur et à
mesure du dialogue multimodal aussi bien en entrée qu’en sortie, permettant aux compagnons virtuels et robotiques d’améliorer leurs capacités communicatives, verbales ou non-verbales, et (2) la dimension multi-parties, c’est-à-dire intégrant plusieurs agents virtuels, robots et humains. Il s’agit de concevoir des modèles et protocoles d’interaction multi-parties, que cette interaction se déroule uniquement entre agents autonomes aussi bien qu’entre agents autonomes et humains dans le cadre de sociétés mixtes. Une façon de générer ces modèles d’interaction est d’utiliser des mécanismes d’apprentissage automatique.
Une des difficultés consistera donc à intégrer des mécanismes dits « just in time » dans un contexte multimodal et multi-parties, permettant la prise en compte de tous les membres d’équipes mixtes agents-humains, dans lesquels les effectifs (nombre d’agents, nombre d’humains dans l’équipe) peuvent varier.
Nous nous focaliserons sur l’utilisation d’ACA et de robots. Dans ce ontexte, la maturation des outils pour la réalité mixte (réalité virtuelle et réalité augmentée), que ce soit en termes logiciel (e.g. Unity3D, Unreal Engine, ARToolKit, …) comme matériel accessible au grand public (Oculus Rift, HTC Vive, Playstation VR, …), et le développement des robots sociaux
(Nao, Pepper, …) permet d’envisager de nouvelles façons d’intégrer ces agents au sein d’environnements mixtes réel/virtuel.

Profil du candidat :
M2 ou école d’ingénieur

Formation et compétences requises :
M2 ou école d’ingénieur

Adresse d’emploi :
INSA Rouen NOrmandie (Saint-Étienne-du-Rouvray)

Document attaché : 202304192002_DescriptionSujet.pdf

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : OpenEdition / LIS
Durée : 3 ans
Contact : patrice.bellot@univ-amu.fr
Date limite de publication : 2023-07-01

Contexte :
Dans le contexte du projet européen COMMONS (https://www.thecommonsproject.org), OpenEdition (https://www.openedition.org) recrute deux doctorant·es pour travailler à concevoir et déployer un observatoire des usages qui permettra de documenter les pratiques des utilisateurs et utilisatrices des quatre plateformes d’OpenEdition, ainsi que d’Huma-Num et Métopes.

Sujet :
Les deux doctorant·es (l’un·e en informatique/humanités numériques et l’autre en sciences de l’information et de la communication/sociologie) collaboreront à la conception d’un protocole d’enquête permettant de documenter les usages des plateformes de science ouverte. Une première campagne d’enquête permettra non seulement un retour critique et réflexif à des fins d’amélioration du protocole mais également une première typologie des usagers et pratiques.

La structure d’accueil du doctorat sera OpenEdition, à Marseille, en co-tutelle avec :
– le LIS (UMR AMU CNRS de l’INS2i) et son école doctorale (ED 184) pour la thèse en informatique ;
– ELICO (Équipe de recherche de Lyon en sciences de l’Information et de la COmmunication) et son école doctorale (485 ED EPIC) pour la thèse en sciences de l’information et de la communication.

Votre CV, un projet de recherche (max. 1 page + 10 références) et une lettre de motivation doivent être déposés sur le portail du CNRS.
Veuillez également mentionner les coordonnées de 2 contacts de référence que nous contacterons directement le cas échéant.

Offre en SIC : https://emploi.cnrs.fr/Offres/Doctorant/UAR2504-SIMDUM-001/Default.aspx
Offre en informatique : https://emploi.cnrs.fr/Offres/Doctorant/UAR2504-SIMDUM-002/Default.aspx

Contact Informatique : Patrice Bellot (AMU CNRS LIS) patrice.bellot@univ-amu.fr ; Simon Dumas Primbault (CNRS OpenEdition) simon.dumas-primbault@openedition.org

Profil du candidat :
Bac +5 informatique (Master Informatique ou équivalent)

Formation et compétences requises :
Apprentissage machine
Recherche d’information
Développement Python

Adresse d’emploi :
Marseille

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : LITIS Lab (Université & INSA Rouen Normandy)
Durée : 3 ans
Contact : paul.honeine@univ-rouen.fr
Date limite de publication : 2023-05-08

Contexte :
Keywords:
Deep learning, anomaly detection, unsupervised learning, optimal transport, domain adaptation, label shift, time series

Supervision Team:
The PhD candidate will be a member of the Machine Learning group in LITIS Lab (University and INSA Rouen Normandy). She/he will be advised by Fannia Pacheco, Paul Honeine, Maxime Berar and Gilles Gasso.

Application:
Please send CV and transcripts of grades to fannia.pacheco@univ-rouen.fr and paul.honeine@univ-rouen.fr.
The deadline for application is the 8th of May 2023.

Sujet :
Description

Deep learning relies on large datasets to learn decision functions for a specific task. These decision functions are prone to be inaccurate at inference for online data, which may be corrupted by anomalies or suffer from a distribution shift. In the most difficult context, training data are labeled, while test data are unlabeled. Under some mild assumptions, the main distribution shift families are covariate shift and label shift. The former is related to causal learning (predicting effects, namely the conditional p(y|x) does not change), and the latter to anti-causal learning (predicting causes, namely the conditional p(x|y) does not change). This thesis focuses on the label shift for online data, since it emerges naturally in diagnosis tasks [1].

The PhD student will take advantage of recent developments in domain adaptation, and more specifically using Optimal Transport (OT), in order to address label shifts in time series. Domain adaptation by OT [2, 3] consists in transporting the source domain feature space to a space equivalent to the target domain space, and then learning a new feature space and decision function where both the source and the target label distributions match. Two major difficulties will be addressed in this PhD thesis. First, we consider unsupervised domain adaptation, namely the target data is available and unlabeled. Whilst label shift is still present in unsupervised domain adaptation, one would like to find the best matching between the source domain and the clusters created in the target domain. To this end, it assumes that all the classes (although unlabeled) are available in the target domain [4, 5]. Second, we consider online domain adaptation, which consists in performing domain adaptation on the fly [6]. This means that the target domain is not available, but new batches of data are available sequentially in order to infer the adaptation, thus requiring online algorithms [7].

This PhD thesis aims to address the most challenging conditions in an online framework and for real-world time-series applications. Its main objectives can be divided into three parts:
i) Study and formalize label shifts and its consequences over classification performance and anomaly detection in times series data.
ii) Propose a method to conceive same label matching in online contexts, by investigating recent advances in OT for domain adaptation.
iii) Create a framework that couples the label matching and unsupervised learning for new distribution discovery.
The proposed framework and devised methods will be evaluated in a variety of time series data for anomaly detection [8], with a focus on fault diagnosis for predictive maintenance in industrial applications [9, 10].

References

[1] Z. Lipton, Y. Wang and A. Smola, “Detecting and correcting for label shift with black box predictors,” ICML, 2018.
[2] N. Courty, R. Flamary, D. Tuia and A. Rakotomamonjy, “Optimal Transport for Domain Adaptation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017.
[3] M. Alaya, M. Berar, G. Gasso, and A. Rakotomamonjy, “Theoretical guarantees for bridging metric measure embedding and optimal transport,” Neurocomputing, 2022.
[4] A. Rakotomamonjy, R. Flamary, G. Gasso, M. E. Alaya, M. Berar, and N. Courty, “Optimal transport for conditional domain matching and label shift,” Machine Learning, 2022.
[5] A. Alaoui-Belghiti et al., “Semi-supervised optimal transport methods for detecting anomalies,” ICASSP, 2020.
[6] M. de Carvalho et al., “ACDC: Online unsupervised cross-domain adaptation,” Knowledge-Based Systems, 2022.
[7] A. Mensch and G. Peyré, “Online Sinkhorn: Optimal transport distances from sample streams,” NeurIPS, 2020.
[8] K. Choi et al., “Deep Learning for Anomaly Detection in Time-Series Data: Review, Analysis, and Guidelines,” IEEE Access, 2021.
[9] F. Pacheco et al., “Deep Ensemble-Based Classifier for Transfer Learning in Rotating Machinery Fault Diagnosis,” IEEE Access, 2022.
[10] P. Honeine, S. Mouzoun, and M. Eltabach. “Neighbor retrieval visualizer for monitoring lifting cranes,” CMMNO, 2018.

Profil du candidat :
The PhD candidate must be a graduate student or have a MSc or engineering degree in one of the following fields: computer science, data science, applied mathematics or equivalent. She/he must have a strong background in machine learning and/or signal processing and/or computer vision. Experience in deep learning is appreciated, as well as proficient programming skills in Python.

Formation et compétences requises :
–

Adresse d’emploi :
Université de Rouen Normadie