Archives

Point cloud based large-scale place recognition (IGN, Paris area)

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

Laboratoire/Entreprise : LaSTIG
Durée : 36 mois
Contact : valerie.gouet@ign.fr
Date limite de publication : 2022-04-08

Contexte :
Thesis proposal: Point cloud based large-scale place recognition – Application to the prevention against fake news

*** Subject of the thesis

The thesis project focuses on 3D point cloud based large-scale place recognition, with the application of geolocation of 3D image data. Without any extra information of the initial position, geolocalazing image content relies on the indexing and retrieval of content similarities in a geolocalized reference. This thesis proposes to study this type of approach by exploiting 3D maps based on acquisition campaigns (in particular LiDAR) that are becoming mainstream thanks to high quality geometry reconstruction which makes them attractive, but also complex to handle given their volume and diversity. Please consult the full text in PDF for the description of the subject thesis.

*** Context

The fields of application of place recognition from images are numerous, we will deal here with the case of the geolocation of amateur video sequences as a certification tool for the prevention against fake news. Massively spread on social networks and on the web, amateur videos relaying information or an event are now very important, with among them content that is fake news, i.e. taken outside of its original context, to express bad or false information. To fight against this form of misinformation, several media, such as the French public television channel “France TV”, have set up a fact checking unit of images and videos which analyzes, verifies and certifies these streams. This complex work is done by hand and would benefit from being automated by using artificial intelligence tools. The verification of geolocation was recognized as essential to best explain what is happening. It is in this collaborative context between IGN and France TV that we focus on this geolocation criterion with the desire to exploit the best georeferencing repositories of today to offer automatic large-scale geolocation solutions, which can, among other things, contribute to the fact checking of visual information.

Sujet :
Full description in English: https://www.umr-lastig.fr/vgouet/News/annonce_these_PlaceReco3D_2022-EN.pdf

Full description in French: https://www.umr-lastig.fr/vgouet/News/annonce_these_PlaceReco3D_2022-FR.pdf

Profil du candidat :
*** Candidate profile

Bac+5 in computer science, applied mathematics or geomatics (master or engineering school).

Please note the only students from the European Union, the United Kingdom or Switzerland are eligible for this thesis project.

*** How to apply

Before March 28, 2022, please send both contacts in a single PDF file the following documents:
– A detailed CV
– A topic-focused cover letter
– Grades and ranks over the last 3 years of study
– The contact details of 2 referents who can recommend you

*** Contacts

– Laurent Caraffa – Laurent.Caraffa@ign.fr, Researcher at LaSTIG (thesis supervisor), IGN, Gustave Eiffel University
– Valérie Gouet-Brunet – Valerie.Gouet@ign.fr, Research director at LaSTIG (director of the thesis), IGN, Gustave Eiffel University

Formation et compétences requises :
A good background in machine learning is required, and a knowledge on 3D computer vision or image indexing will be appreciated. The successful candidate must have good programming skills (Python, C/C++). Although fluency in French is not required, fluency in English is necessary. Curiosity, open-mindedness, creativity, perseverance and the ability to work in a team are also key personal skills in demand.

Adresse d’emploi :
*** Organization

* Start: last quarter of 2022

* Place: the thesis will be carried out in Paris area at the LaSTIG laboratory, located in Saint-Mandé (73 avenue de Paris, Saint-Mandé metro, line 1) in the premises of the IGN. The doctoral student will be attached to the MSTIC Doctoral School (ED 532).

The French mapping agency IGN (National Institute for Geographic and Forest Information) is a public administrative establishment attached to the French Ministry of Ecological Transition; it is the national reference operator for mapping the French territory. The LaSTIG Laboratory in Sciences and Technologies of Geographic Information for the smart city and sustainable territories, is a joint research unit attached to the Gustave Eiffel University, the IGN and the School of Engineering of the city of Paris (EIVP). It is a unique research structure in France and even in Europe, bringing together around 80 researchers, who cover the entire life cycle of geographic or spatial data, from its acquisition to its visualization, including its modeling, integration and analysis; among them about thirty researchers work in image analysis, computer vision, machine learning, photogrammetry and remote sensing. LaSTIG researchers can be involved in the teaching activities of the IGN engineering school, the ENSG (Ecole Nationale des Sciences Géographiques), which offers access to undergraduate and graduate students with excellent quality in fields related to geographic information sciences: geodesy, photogrammetry, computer vision, remote sensing, spatial analysis, cartography, etc.

Document attaché : 202202212218_annonce_these_PlaceReco3D_2022-EN.pdf

MCF 27ème section en délégation à l’Université de la Nouvelle Calédonie

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

Laboratoire/Entreprise : Institut des Sciences Exactes et Appliquées (ISEA)
Durée : 2ans renouvelable
Contact : nazha.selmaoui@univ-nc.nc
Date limite de publication : 2022-04-08

Contexte :
Recrutement d’un MCF en délégation à l’Université de la Nouvelle Calédonie.

Sujet :
voir le profil dans le document joint.

Profil du candidat :
La personne recrut e aura un profil recherche lié à l’apprentissage machine, la fouille de données, la science de données, le big data et applications, ainsi qu’une polyvalence en ce qui concerne l’enseignement.

Formation et compétences requises :
être déjà titulaire dans l’enseignement supérieur.

Adresse d’emploi :
Nouméa, Nouvelle Calédonie

Document attaché : 202202212158_MCF-CNU-27-D-prolongation-28.02.2022.pdf

Handling classes’ imbalance in supervised classification for medical diagnostics

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

Laboratoire/Entreprise : LAMADE – Pôle Sciences des Données – Université P
Durée : 5-6 months
Contact : sana.mrabet@dauphine.psl.eu
Date limite de publication : 2022-03-05

Contexte :
The classification of highly imbalanced data is a big challenge for machine learning techniques. To deal with this challenge, many solutions have been proposed that could be classified in three categories: data pre-processing with under/oversampling technique that creates a training sample with a new instances distribution, active sampling that changes the training sampling throw the learning process, and the Synthetic Minority Over-sampling Technique (SMOTE) that creates new synthetic instances in the minority class. The efficiency of each approach depends on the context. For the medical diagnostics, if the input data contains categorical attributes, the SMOTE methods could be not suitable. Otherwise, if the data imbalance ratio is high, using the under/oversampling could induce loss of information in the training sample

Sujet :
Study and compare three different approaches to handle classes’ imbalance in medical data: data pre-processing with over/under sampling, synthetic minority over-sampling and active sampling.

Profil du candidat :
Master 2 ou dernière année d’école d’ingénieur en informatique

Formation et compétences requises :
Bonne connaissance en Machine Learning et en programmation Python.
Maîtrise de l’anglais et bonne capacité rédactionnelle

Adresse d’emploi :
Université Paris Dauphine – PSL
Place du Maréchal de Lattre de Tassigny – 75775 PARIS Cedex 16

Document attaché : 202202211348_Proposition sujet mémoire 2022.pdf

Deep neural network compression using tensor methods

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

Laboratoire/Entreprise : laboratoire d’informatique et systèmes (LIS) UMR
Durée : 5 to 6 months
Contact : zniyed@univ-tln.fr
Date limite de publication : 2022-04-30

Contexte :
Deep Neural Networks (DNNs) demonstrate good prediction performances in numerous applications. However, the architectures of neural networks are very large, reaching several million parameters, and running them on systems with limited computing capacity (embedded systems) becomes a difficult task. For this reason, we will focus in this internship project on the compression of DNNs by tensor methods.

Sujet :
This internship project deals with the study of new compression techniques for deep neural networks, by resorting to tensor decompositions to model and factorize the DNN weights. Recent studies show that DNN weight matrices are often redundant, and by restricting their ranks, it is possible to significantly reduce the number of parameters without a significant drop in performance. In this project, we propose to convert these matrices to a tensorial format and to use multidimensional data processing methods to compress them. The goal of this internship is to study different tensor representations, such as the canonical polyadic decomposition (CPD) or Tucker decomposition (TD), for the compression of the converted multidimensional weights. Specifically, we will study the compactness of these representations and their impact on the predictive accuracy of DNNs. In a first stage, the intern student will review the existing state-of-the-art tensor-based compression techniques and will get familiar with the tensor decompositions. Then, we will compare different representations with the goal to improve them and propose new tensor-based scheme for DNN compression.

This internship can be followed by a Ph.D research project, starting October, 2022, at LIS, Toulon

Profil du candidat :
M2R or engineering school students with major in signal processing, machine learning or applied mathematics.

Formation et compétences requises :
Good python programming skills are required. The knowledge of deep learning frameworks is a desirable plus. The candidate should have good writing and oral communication skills.

Adresse d’emploi :
The intern student will join the Signal and Image (SIIM) research team at the LIS laboratory, Toulon.
The internship will be supervised by Yassine Zniyed (Associate Professor at Université de Toulon) and Thanh Phuong Nguyen (Associate Professor/HDR at Université de Toulon).

Document attaché : 202202210815_Stage_M2R_2022.pdf

1st call-for-participation JOKER@CLEF: Automatic Wordplay and Humour Translation Task

Date : 2022-04-22

Deadlines

Data & guidelines release: February – March 2022

Run submission: 22 April 2022

Draft paper submission: 27 May 2022

CLEF conference: 5–8 September 2022

Context

Humour remains one of the most difficult aspects of intercultural communication: understanding humour often requires understanding implicit cultural references and/or double meanings, and this raises the question of its (un)translatability. Wordplay is a common source of humour due to its attention-getting and subversive character. The translation of humour and wordplay is therefore in high demand. Modern translation depends heavily on technological aids, yet few works have treated the automation of humour and wordplay translation, or the creation of humour corpora. The goal of the JOKER workshop is to bring together translators and computer scientists to work on an evaluation framework for wordplay, including data and metric development, and to foster work on automatic methods for wordplay translation.

Tasks

We invite you to submit both automatic and manual runs! Manual intervention should be reported.

Task 1: Сlassify and explain instances of wordplay.

Task 2: Translate single words containing wordplay.

Task 3: Translate entire phrases containing wordplay.

Unshared task: We welcome any other type of submission that uses our data as an open task.

How to participate
Sign up at the CLEF website (https://clef2022-labs-registration.dei.unipd.it/). All team members should join the JOKER mailing list (https://groups.google.com/u/4/g/joker-project). The data will be made available to all registered participants.

Contacts

JOKER website: http://joker-project.com/

CLEF website:
https://clef2022.clef-initiative.eu/index.php

Registration: https://clef2022-labs-registration.dei.unipd.it/

Email: contact@joker-project.com

Twitter: https://twitter.com/joker_research

Google Group: https://groups.google.com/u/4/g/joker-project

Notre site web : www.madics.fr
Suivez-nous sur Tweeter : @GDR_MADICS
Pour vous désabonner de la liste, suivre ce lien.

1-st Call for Participation – SimpleText Track @ CLEF-2022

Date : 2022-04-22

Context
The web and social media have democratized information sharing and have become the main source of information for citizens, risking users to rely on shallow information in sources prioritizing commercial or political incentives rather than the correctness and informational value. Non-experts tend to avoid scientific literature due to its complex language, internal vernacular, or their lack of prior background knowledge. Text simplification approaches hold the promise to remove some of these barriers. The SimpleText track is a part of the CLEF initiative which promotes the systematic evaluation of information access systems, primarily through experimentation on shared tasks. SimpleText addresses the challenges of text simplification approaches in the context of promoting scientific information access, by providing appropriate data and benchmarks. The track uses a corpus of scientific literature abstracts and popular science requests. It features three tasks.

Tasks
We invite you to submit both automatic and manual runs! Manual intervention should be reported.

* Task 1: What is in (or out)? Select passages to include in a simplified summary, given a query.

* Task 2: What is unclear? Given a passage and a query, rank terms/concepts that are required to be explained for understanding this passage (definitions, context, applications,..).

* Task 3: Rewrite this! Given a query, simplify passages from scientific abstracts.

* In addition, we welcome any other type of submission that uses our data as an open task.

How to participate
In order to participate, you should sign up at the CLEF website (https://clef2022-labs-registration.dei.unipd.it/). All team members should join the SimpleText mailing list (https://groups.google.com/g/simpletext). The data will be made available to all registered participants.

Deadlines

*Data release: February 2022

*Final guidelines: March 2022

*Run submission: 22 April 2022

*Results available: 6 May 2022

*Draft paper submission: 27 May 2022

*Camera-ready: 1 July 2022

*CLEF conference: 5-8 September 2022

Lien direct

Notre site web : www.madics.fr
Suivez-nous sur Tweeter : @GDR_MADICS
Pour vous désabonner de la liste, suivre ce lien.

Journée d’étude Mots/Machines #4: Simplification et adaptation du texte

Date : 2022-04-22
Lieu : Université de Bretagne Occidentale

20, rue Duquesne – CS9383729238

Brest Cedex 3, France

La simplification de textes est utilisée dans les domaines de la traduction,la localisation et la rédaction technique. La pré-édition consiste à préparer un document avant d’appliquer la traduction automatique afin d’obtenir de meilleurs résultats et de réduire la charge de travail de post-édition. C’est devenu un choix courant pour les entreprises publiant leur contenu dans différentes langues.

En outre, la simplification du texte améliore les applications de traitement automatique de la langue naturelle, notamment les résultats de la traduction automatique. Ainsi, la simplification automatique de textes pourrait s’avérer utile dans divers domaines tels que la communication scientifique, le journalisme scientifique, la politique et l’éducation, tant pour les cours de sciences que pour la didactique. La vulgarisation scientifique et le journalisme scientifique sont d’ailleurs l’un des plus anciens programmes de l’UNESCO.

Les textes simplifiés sont également plus accessibles aux locuteurs non natifs, aux jeunes lecteurs, aux personnes souffrant d’un handicap de lecture ou ayant un niveau d’éducation inférieur (objectif de développement durable INÉGALITÉ RÉDUITE).

Les textes scientifiques, tels que les publications de recherche, peuvent être difficiles à comprendre pour les non-experts du domaine ou les scientifiques qui ne sont pas concernés par la publication. L’amélioration de la compréhensibilité des textes et leur adaptation à différents publics restent un problème non résolu. La simplification de textes est un pas en avant vers la recherche réellement ouverte, accessible et compréhensible par tous, le développement d’un contre-discours aux fake news basées sur des résultats scientifiques, la possibilité s de lire plus rapidement et par conséquent, de devenir mieux informé.e sur les résultats scientifiques, notamment avec l’explosion de la science ouverte depuis le début de la pandémie actuelle de COVID-19 (objectif de développement durable ÉDUCATION DE QUALITÉ).

L’objectif de cette journée d’étude est de fournir une plateforme de communication à une communauté interdisciplinaire de chercheurs en traduction, rédaction technique, traitement du langage naturel, recherche d’information, linguistique, didactique, journalisme scientifique et vulgarisation scientifique.

Lien direct

Notre site web : www.madics.fr
Suivez-nous sur Tweeter : @GDR_MADICS
Pour vous désabonner de la liste, suivre ce lien.

Discovery Science 2022

Date : 2022-04-22
Lieu : Montpellier, France

The 25th International Conference on Discovery Science (DS 2022)

https://ds2022.sciencesconf.org/

Montpellier, France, October, 10-12, 2022

***********************************
COVID-19
***********************************
We hope that by October the world will have returned to normality and we can welcome you in Montpellier. However, in case the COVID-19 risk persists and traveling is difficult, DS 2022 will take place either as a mixed event by offering both remote and on site presentation options or as a fully online event in the worst case. The accepted papers will still be published by Springer and the special issue will proceed as announced. In these challenging times that the whole of humanity is going through, we hope that all of you are safe and remain healthy.

***********************************
::: Scope :::
***********************************
The 25th International Conference on Discovery Science (DS 2022) provides an open forum for
intensive discussions and exchange of new ideas among researchers working in the area of Discovery Science. The scope of the conference includes the development and analysis of methods for discovering scientific knowledge, coming from machine learning, data mining, intelligent data analysis, big data analysis as well as their application in various scientific domains.

***********************************
::: Submission Topics :::
***********************************
We invite submissions of research papers addressing all aspects of discovery science: papers that focus on the analysis of different types of massive and complex data, including structured, spatio-temporal and network data. We would also like to encourage contributions from the areas of computational scientific discovery, mining scientific data, computational creativity and discovery informatics.
We particularly welcome papers addressing applications from different domains of science including biomedicine and life sciences, astronomy, physics, chemistry, as well as social sciences. Applications to massive, heterogeneous, continuous or imprecise data sets are of particular interests. Possible topics include, but are not limited to:

Knowledge discovery, machine learning and statistical methods
Ubiquitous Knowledge Discovery
Data Streams, Evolving Data and Models
Change Detection and Model Maintenance
Active Knowledge Discovery
Information extraction from scientific literature
Knowledge discovery from heterogeneous, unstructured and multimedia data
Data and knowledge visualization
Planning to Learn
Knowledge Transfer
Computational Creativity
Human-machine interaction for knowledge discovery and management
Evaluation of models and predictions in discovery setting
Causality modelling
AutoML, meta-learning, planning to learn
Explainable AI, interpretability of machine learning and deep learning models
Learning from complex data
Graphs, networks, linked and relational data
Spatial, temporal and spatiotemporal data
Unstructured data, including textual and web data
Multimedia data
AI frameworks for discovery in scientific domains
Biomedical knowledge discovery, analysis of (multi)omics, micro-array, gene deletion, gene set enrichment data
Machine Learning for High-Performance Computing, Grid and Cloud Computing
Applications of the above techniques in scientific domains, such as
Physical sciences (e.g., materials sciences, particle physics)
Life sciences (e.g., systems biology/systems medicine)
Environmental sciences
Life Sciences
Natural and social sciences

***********************************
::: Publishing :::
***********************************
Traditionally the proceedings of DS series appear in the Lecture Notes in Artificial Intelligence Series by Springer-Verlag. In addition, authors of best papers will be invited to submit their extended versions to a special issue on Discovery Science of the Machine Learning journal published by Springer. Fast Track Processing will be used to have them reviewed and published.

***********************************
::: IMPORTANT DATES :::
***********************************
Abstract submission: May 23, 2022
Full paper submission: May 30, 2022
Notification: July 20, 2022
Camera ready version, author registration: August 8, 2022
Conference: October 10-12, 2022

***********************************
::: Submission guidelines :::
***********************************
Regular research papers may contain up to fifteen (15) pages and must be formatted according to the layout supplied by Springer-Verlag for the Lecture Notes in Computer Science series. The Program Committee reserves the right to offer acceptance as Short Papers (10 pages in the Proceedings) to some submissions. The reviews are single-blind. You do not need to anonymize your submission.
Submitted papers may not have appeared in or be under consideration for another workshop, conference or a journal, nor may they be under review or submitted to another forum during the DS 2022 review process.
We encourage all authors to include their individual ORCID in their address information.
Authors can submit their regular papers via our submission page through Easychair:

https://easychair.org/my/login_author?sum=073323801fd3b7125c2b6cc57ecf0a6f;conference=267691

Authors of accepted papers must submit along with the final version of their paper a consent to publish, filled and signed. Authors of accepted papers are expected to register to the conference and present their work (see author registration date).

***********************************
Special issue and Best Student Paper Award
***********************************
The authors of a number of selected papers presented at DS 2022 will be invited to submit extended versions of their papers for possible inclusion in a special issue of Machine Learning journal (published by Springer) on Discovery Science. Fast-track processing will be used to have them reviewed and published.
There will be an award for the Best Student Paper in the value of 555 Euro sponsored by Springer.

Dino Ienco (PC Co-Chairs DS)
Pascal Poncelet (PC Co-Chairs DS)
Sašo Džeroski (General Chair DS)

Lien direct

Notre site web : www.madics.fr
Suivez-nous sur Tweeter : @GDR_MADICS
Pour vous désabonner de la liste, suivre ce lien.

Apprentissage profond sur données synthétiques appliqué à l’imagerie radar

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

Laboratoire/Entreprise : ONERA / UTT
Durée : 3 ans
Contact : alexandre.baussard@utt.fr
Date limite de publication : 2022-05-02

Contexte :
voir fichier attaché

Sujet :
voir fichier attaché

Profil du candidat :
voir fichier attaché

Formation et compétences requises :
voir fichier attaché

Adresse d’emploi :
ONERA et Université de Technologie de Troyes

Document attaché : 202202171418_phy-demr-2022-11.pdf

PhD in Computer Sciences / Computational Social Sciences

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

Laboratoire/Entreprise : LPI/CRI-Paris, Université Paris Cité
Durée : 3 ans
Contact : pedro.ramaciottimorales@learningplanetinstitute.org
Date limite de publication : 2022-03-12

Contexte :
The Learning Planet Institute (LPI) is an interdisciplinary research unit of Université Paris Cité, developing diverse projects in themes ranging from systems biology to network sciences and complex systems. In the heart of Paris, the LPI brings together social scientists, biologists, designers, computer scientists, mathematicians and physicists among other disciplines, to develop research seeking high societal impact.

Sujet :
This position is part of an initiative to investigate the challenges wrought on democracies by the Internet and Artificial Intelligence, and to improve the understanding of the impact they have in society. The goal of the initiative is to improve the understanding and interpretability of AI systems that mediate social public life in social networks, media platforms, and online news outlets. How do Recommender Systems perceive and model the digital landscape of users and contents to recommend us friends and information? What is the relation between algorithmic recommendations mediating the activity in large internet platforms and the social phenomena such as echo chambers and polarization? This initiative relies on mathematical modeling, political science survey data, and computational experiments with Recommender Systems to develop actionable theories of machine social cognition and tool kits to analyze models learned and leveraged by AI architectures.

Profil du candidat :
The hired doctoral researcher will conduct data analyses of social and media platform data and theoretical modelization work. It is also expected that the doctoral researcher will conduct experiments, training models, and develop software tools to further the understanding of AI systems and their social cognition.

We encourage students with a background in natural sciences and technology (e.g., engineering, computer science, mathematics, physics) to apply for the position. Applicants with different backgrounds and strong modeling and computing skills are also encouraged to apply.

Formation et compétences requises :
Experience with Machine Learning in Python.
Interest in learning big data technologies.
Interest in doing research in AI interpretability.
Experience/interest in working in research in mathematical modeling (geometrical modeling of learning space for Deep Learning).
Interest in working with interdisciplinary teams in a public policy-inspired environment.

Adresse d’emploi :
CRI-Paris
8 bis Rue Charles V, 75004 Paris

Document attaché : 202202170855_Fiche de poste doctoral student LPI.pdf