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

Laboratoire/Entreprise : Laboratoire Hubert Curien, University of Saint-Eti
Durée : 3 years
Contact : amaury.habrard@univ-st-etienne.fr
Date limite de publication : 2022-05-09

Contexte :
A Phd position covering Deep Learning, color Science, Photonics and Laser Processing is available at the University of Saint-Etienne

Sujet :
Laser processing is a flexible and cost-effective tool that recently opened new perspectives of applications in industry. Implemented on plasmonic metasurfaces, it produces very singular colors that can be tuned independently in different modes of observation. Laser-induced plasmonic colors thus enable printing multiplexed images, which have great promise in security printing and data storage1,2. However, the latter require very good accuracy in color printing. And, laser induced colors strongly depend on the initial state of the material. Predicting the full gamut of colors that can be observed in different modes of observation on plasmonic metasurfaces processed by a large set of laser processing parameters, when the initial state of these metasurfaces can vary from one batch to another, appears then as a crucial step for industrial implementation. As physical models are missing for such predictions, other approaches must be found.
Deep learning represents one of the most powerful family of models in machine learning when one has to make some predictions from data having some local structure such as images or surfaces. In this thesis, the objective is to provide some appropriate architectures accompanied with relevant objective functions to correctly train these architectures for accurate prediction of the laser printed colors3. A first challenge is to take into account physical properties of the materials and the laser processing parameters in the model. Then, another goal is to improve the robustness of the model by adapting existing adversarial robustness methods existing in image classification to laser printed colors4. Finally, the third aspect tackled in this project is to develop models for being able to automatically adapt the learned models to slightly different initial metasurfaces by means of transfer learning/domain adaptation strategies. This last objective intends to offer a certain tolerance to unwanted variations in the initial metasurface elaboration while maintaining a very good accuracy on the prediction of the laser printed colors.
References
(1) N. Destouches, et al. “Laser-empowered metasurfaces for white light image multiplexing”, Adv. Func.
Mater., 31, 2010430 (2021)
(2) N. Dalloz, et al. “Anti-counterfeiting white light printed image multiplexing by fast nanosecond laser
processing”, Adv. Mater., 34, 2104054 (2021)
(3) M. Raissi. “Deep hidden physics models: Deep learning of nonlinear partial differential equations”,
Journal of Machine Learning Research, 19(25):1–24, 2018.
(4) A. Shafahi, et al. “Adversarially robust transfer learning”, in International Conference on Learning
Representations (ICLR), 2020.

Applications must be sent to both contact email addresses as soon as possible and before May 1st, 2022. Application can be written in French or English.
All applications must contain:
-a CV, with a possible list of publications and conferences, and the CEFR level in English (except if university courses were taught in English)
-a short motivation letter explaining why you should be successful in this
research work
-Bachelor degree and transcripts
-Master transcripts (at least semester 1 and 2. Semester 3 if available)
-An example of project that you have carried out in deep learning
-References of academics to be contacted (or recommendation letters).

Application deadline: May 1st, 2022

Start of the position: October 1st, 2022

Contacts: Prof. Nathalie Destouches: nathalie.destouches@univ-st-etienne.fr, and Prof. Amaury Habrard: amaury.habrard@univ-st-etienne.fr

Link:
https://laboratoirehubertcurien.univ-st-etienne.fr/en/teams/functional-materials-and-surfaces/job-opportunities/phd-offer-prediction-of-multidimensional-colors-printed-by-laser-on-plasmonic-metamaterials-using-deep-learning-and-adaptive-strategies.html

Profil du candidat :
Master of Science in Computer Science or in Physics

Formation et compétences requises :
-Very good knowledge and experience in machine learning and deep learning
-Background in color science and image processing
-Some knowledge in photonics and eager to expand his/her experimental skills in this field
-Open-minded, curious and interested in working with both computer scientists and physicists
-Ability to take initiatives and work in autonomy

Adresse d’emploi :
Campus manufacture
Laboratoire Hubert Curien, University of Saint-Etienne
Saint-Etienne, France

Document attaché : 202203101326_PhD offer 2022.pdf

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

Laboratoire/Entreprise : LISTIC
Durée : 3 ans
Contact : yajing.yan@univ-smb.fr
Date limite de publication : 2022-07-10

Contexte :
This Ph.D thesis is proposed along with the increasing and regular availability of the amount of remote sensing data and the response to the requirement of operational prediction of natural hazards. The main objective is to improve the near-real-time integration of remote sensing data and dynamical geophysical models for the mitigation of natural hazards. This thesis is partly (50%) funded by the national action plan in Artificial Intelligence. The relevance of the methodology developed in this thesis 1) compared to the actually emmerging data-driven methods, lies in the incorporation of geophysical knowledge (which helps increase the interpretability and the accountability of the results for operational purpose) and its near-real-time implementation ; 2) compared to previous attemps to improve the near-real-time integration of InSAR data based on
the Kalman Filter, lies in the capability in taking non-Gaussian error statistics (which can represent better the reality) into account. First application will be in volcanology, using InSAR & GNSS data, but the methodology can be easily utilized for other natural hazards (e.g. landslides, slow slip, etc.), as well as for anthropogenic hazards like forest fire.

Sujet :
In a perspective of volcanic hazard assessment, it is fundamental to be able to know, in advance, if magma that has started to propagate from a reservoir will reach the surface, where and when. The propagation phase is generally rapid, lasting a few hours to a few months but it induces
seismicity and deformation signals. These signals are recorded by continuous sensors (GNSS) and InSAR data whose temporal sampling frequency as well as latency have been greatly improved in recent years. In addition, we have dynamic magmatic intrusion propagation models that can be used to calculate, depending on the physical parameters of the Earth’s crust, the properties of the magma and the state of local stress, the trajectory followed by the magma and its propagation velocity (Pinel et al. 2017). Data assimilation, a method that combines a dynamic model with observations at present and in the past based on error statistics and predicts the future state of the observed system, is therefore an appropriate tool to respond to the need to be able to predict
the position and timing of an eruption in volcanology.

Among numerous data assimilation methods, the particle filter is distinguished from others by its great ability to deal with non-linear models and non-Gaussian error statistics (van Leeuwen P.J 2009, van Leeuwen P.J, 2010). The particle filter is based on a representation of the probability density of the dynamic model by a discrete set of states of the model (namely particles) and relies on the Bayes theorem so without worrying about the distribution of model errors (different from most other data assimilation methods which assume a Gaussian distribution of the errors). The evolution of the probability density of the model is realized through the propagation of all the particles (states of the model) over time following the model equation (the physics). When
observations of the system are available, the relative contributions of the particles are modified so that the information available in the observations is integrated into the particles swarm. The particle filter does not directly correct the values of particles, but their relative contributions, this is very important for estimating magmatic intrusion propagation trajectories. The particle filter is therefore the appropriate tool in the present specific context of estimating the position of a volcanic eruption.

In this thesis, we will develop an efficient data assimilation strategy using the particle filter allowing to use timely available geodedic data to predict the location and timing of eruptive vents induced by magmatic intrusion propagation. This work will be based on the Ph.D thesis of Mary Grace Bato who, under our supervision and for the first time, successfully applied sequential data assimilation techniques (i.e. Ensemble Kalman Filter) to volcanological problems focussing on the pressurization and rupture of magmatic reservoirs (Bato et al, 2017, Bato et al., 2018). Moreover, this thesis will benefit from the results obtained in the TOSCA AssimSAR project (2018-2019). It will be part of the Franco-German ANR MagmaPropagator (ANR-18-CE92-0037, 2019-2022) with an application to Piton de la Fournaise volcano. It will also be the oppotunity to collaborate with Andy Hooper from the University of Leeds, who is developping new methods to automatically extract a physical signal from InSAR time series (Gaddes et al, 2019).

Selected references :
1) Bato M.-G., Pinel V., Yan Y., Jouanne F., Vandemeulebrouck J., “Possible deep connection between volcanic systems evidenced by sequential assimilation of geodetic d”, Scientific Reports, Nature Publishing Group, 2018, https://doi.org/10.1038/s41598-018-29811-x
2) Bato M.-G., Pinel V., Yan Y., “Assimilation of Deformation Data for Eruption Forecasting: Potentiality Assessment Based on Syntheti”, Frontiers in Earth Science, Frontiers Media, 2017, pp.doi: 10.3389/feart.2017.00048
3) Dalaison, M., Jolivet, R., A Kalman filter time series analysis method for InSAR, Journal of Geophysical Research : Solid Earth , 2020, e2019JB019150. e2019JB019150 2019JB019150.
4) Gaddes, M. E., Hooper, A., Bagnardi, M. (2019), Using machine learning to automatically detect volcanic unrest in a time series of interferograms, Journal of Geophysical Research : Solid Earth , 124(11), 12304–12322.
5) Pinel V., Carrara A., Maccaferri F., Rivalta E., Corbi F., A two-step model for dynamical dike propagation in two dimensions: Application to the July 2001 Etna eruption, 2017, Journal of Geophysical Research, vol. 122, doi:10.1002/2016JB013630.
6) van Leeuwen P.J., Review Particle Filtering in Geophysical System, Mathematical Advances in Data Assimilation, 2009, pp. 4089-4114.
7) van Leeuwen P.J. Nonlinear data assimilation in geosciences : an extremely efficient particle filter, 2010, Quarterly Journal of the Royal Meteorological Society, vol. 136, pp. 1991-1999.

Profil du candidat :
The Ph.D candidate should have good skills in signal/image processing,
mathematics/statistics or geophysics.

Formation et compétences requises :
mathematics/statistics, image processing, remote sensing

Adresse d’emploi :
Laboratoire d’Informatique, Systèmes, Traitement de l’Information et de la Connaissance, Université Savoie Mont-Blanc, Annecy, France

Document attaché : 202203081350_sujet_these_DA_volcan_en2022.pdf

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

Laboratoire/Entreprise : LAMSADE
Durée : 3 years
Contact : sana.mrabet@dauphine.psl.eu
Date limite de publication : 2022-04-15

Contexte :
Evaluating the level of biodiversity in marine and terrestrial locations is currently a universal practice for estimating the impact of human activities and climate change on the ecosystems of these environments. An innovative and modern approach to perform this observation is the high-throughput sequencing of DNA taken from the environment (“environmental DNA)
Environmental DNA (eDNA) analysis consists mainly of the identification of species from the DNA they leave in their environment. Many studies show that the use of eDNA allows a good estimation of the different species (or taxa) present in different types of environments. eDNA analysis is based on classical molecular biology techniques (PCR, sequencing…). The identified taxa are assigned to ecological weights which are used to calculate the biotic indices (BI). These indices are used to determine the ecological quality of the site in question (generally classified in five categories from “very poor” to “very good”). However, these studies relied on reference sequence databases (Silva Greengenes LTP) for taxonomic assignment, in order to retrieve taxon-specific ecological weights and to calculate BI values. This phase may fail if some sequences in the environment to be analysed are incomplete or do not appear in the reference databases.
Recent works [1,2,5] has demonstrated that machine learning can be used to predict accurate values of biotic indices from eDNA metabarcoding, regardless of sequence affiliation. The idea is to generate a clustering model to group closely related sequences (high percentage of identity) belonging to a well-defined taxon. The approach then consists in using a clustering technique on the analysis data of the marine samples applicable even if the determined sequences are incomplete. These sequences are then affiliated automatically to the taxon of the central sequence of the cluster identified in a reference database. The validity of the model can be checked by searching for the exact affiliation of the cluster centres in reference databases.

Sujet :
The subject of this thesis has a double objective. The first objective is the implementation of conventional techniques for clustering eDNA data of a marine environment allowing to evaluate the BI of this environment. The generated models in this phase may be non-generalizable solutions and very dependent on the sequencing methods used in the analysed samples. Hence the second objective of the thesis which aims at the generalization of prediction models to data from other samples of the same marine environment using evolutionary methods. The idea is to develop a hybrid evolutionary algorithm that evolves clustering models as a function of sample parameters (i.e. sequencing markers) and clustering parameters (i.e. distance measurement indicators). The model scheme can be based on the “Genetic K-means algorithm” (GKA) of Krishna and Murty.
The first models are to be tested on marine samples studied by Cordier [1,2] and available online. Other samples will be studied after validation of the algorithm. The validation of the method and its application on other marine data may lead to a possible exchange with OFB (French Office of Biodiversity).

Profil du candidat :
Master’s degree in computer science or equivalent degree giving access to PhD studies.

Formation et compétences requises :
– Good knowledge in Machine Learning
– Good knowledge in Computer Science and Mathematics.
– Good programming skills, especially in Python programming.
– Some knowledge in Bio-informatic would be good but it is not mandatory.
– Good command of written and spoken English.

Adresse d’emploi :
LAMSADE CNRS UMR 7243, University of Paris Dauphine, France

Document attaché : 202203072123_PhD Application – Lamsade 2022.pdf