Open position for post-doc and engineer: Smart cities and fog computing Laboratory: IRIT (Computer science Laboratory) in Toulouse in the SEPIA TEAM Localisation: University Toulouse III Salary: Starts at 2650€ brut for post-doc and 2330€ for Engineer and increases depending on previous experience.

Duration: 1 year, can be extended for another year Start of the position: September 2022 Keywords: Edge and Fog computing, Gama, Multi-agent systems, scheduling

Profiles: Master or PhD Expected abilities, one or more of the following Distributed systems Optimization techniques (A.I, Multi-Agent System) Graph theory Fog/Edge computing Knowledge on Gama platform is a plus

Context

The project is in the context of the Vilagil (funded by Toulouse Metropole) project at IRIT (http://www.irit.fr) in Toulouse. The host team is SEPIA. The main research topic of the SEPIA team is optimisation of datacenter (multi-objective scheduling). This project aims at improving mobility using a smart-city approach.

A large amount of local information will be harnessed to help proposing to citizens the best paths in the city using multiple mobility means (bikes, buses, metro, cars, ...).

The computing infrastructure backing this service is currently centralized, which leads to several type of problems: from a high resource consumption to a single point of failure.

In the context of smart cities, computing infrastructures are geographically distributed.

In our project we will use fog and edge computing to manage distributed applications providing the same service as the centralized one, and running on fixed and mobile entities.

The scientific aim of our project consists in mapping the graph of the physical infrastructure with the one of the services to be deployed.

Assignement The goal of this project will be to develop algorithms to place multi-modality mobility services on the Fog infrastructure of Toulouse Metropole. We consider having a graph of a complex application with sources of information (sensors in the city) and treatments (such as requests of path from end-users).

We also have a graph of physical computing infrastructure available.

We will use both of them to optimize the services deployment under different constraints and objectives (such as Quality of Service, Energy Consumption, ...) and to provide a simulation and visualisation platform of the results.

We will take into account the dynamism of the system (such as people and bus mobility, different densities of requests, ...). This position aims at: Formalizing the problem; Explore different possible heuristics; Develop a simulator for evaluation and visualisation based on Gama simulator.

Depending on the applicant profile (toward research or engineering) the content of the position can be amended.

Application You can submit your application (CV/Cover Letter) to Georges Da Costa ([email protected]) and Patricia Stolf ([email protected])

A 30 month position is open at CEA around AI Safety. You can apply through the following link (french version)

https://www.emploi.cea.fr/-job-offer/job-engineer-ia-of-confidence-h-f_22074.aspx? LCID=1036

There is an english version but the translation of the job description and profile were mixed up with another announcement. They will soon be replaced with the following:

Job description You will participate in the development of one of the laboratory's tools, PyRAT (Python Reachability Analysis Tool), for the verification of safety and security properties of artificial intelligence-based systems using formal methods such as abstract interpretation. This will include contributions to:

- the various actions related to PyRAT in the framework of the Confiance.AI project. - the development of new features for the PyRAT tool and extend the support of existing features.

- the study of the state of the art around new artificial intelligence techniques and their verification as well as to the evaluation of their potential contribution and support by PyRAT. - the elaboration of new methods to verify AI systems and study their contribution to PyRAT.

- the development of other tools related to PyRAT or PyRAT front-end such as AIMOS or CAISAR and their integration with PyRAT. - the improvement the visualization features of PyRAT to help developing the tool's GUI.

The safety and security of AI-based software is a key objective for new industrial systems, both critical and non-critical, which increasingly rely on AI techniques. Traditional software verification provides a base of methods and experiences in the field of formal verification that must now be applied to AI.

More generally, AI systems are becoming more and more diversified with recent advances such as XGBoost networks, Transformers, Recurrent Nets, ... adding to older methods already diverse not currently treated such as decision trees. It is to this increasing diversification that we must face in order to succeed in ensuring the safety and security of these new systems.

Applicant Profile - Engineer or master of science - AI basics/experience (Tensorflow/Pytorch/Keras) - Python development skills - knowledge for GUI (ReactJS or other) - (optional) notions of abstract interpretation/formal methods

Title: Urban Knowledge Hub for Evolving Cities

Supervisors: Dr. Emmanuel Coquery, Dr. John Samuel, Prof. Gilles Gesquière

University: Université Lyon 1 (UCBL), LIRIS Laboratory (CNRS UMR 5205)

Date of recruitment: 1 September 2022

Duration: 36 months (3 years)

Keywords: Knowledge representations, multidimensional data, data evolution, code evolution, knowledge evolution, reproducibility

Context

This project proposal for a Ph.D. position has been selected in the IADoc@UdL call. The goal is about thinking about how to capitalize on the amount of knowledge developed during the last decade and use it in a multidisciplinary context for understanding city evolution and its capacity to become more sustainable and resilient. This proposal is made possible thanks to a strong collaboration between LIRIS Laboratory and Metropole of Lyon. This project is fully integrated within the LIRIS laboratory where it brings the necessary transversal approach to talk about smart cities between data science specialists (BD team) and graphic computing specialists (Origami team) since it is a question of keeping the strong link between vector data and associated semantics. The LIRIS is also a key partner when it comes to working in the field of Artificial Intelligence (AI).

As an element of context, research is not just about publications and providing code. Working in an urban context implies rethinking and discovering the unexplored part of knowledge (data, code, literature, and process) like exploring the submerged part of an iceberg. In this thesis, the major theme is how to propose a real aggregation of this knowledge about the city, an aggregation made in a diachronic context?

In this proposal, we seek a candidate, with good knowledge of computer sciences (level Bac+5 in France). The position is available in Lyon (campus LyonTech La Doua). More information on this subject is given below.

More information on this subject

Evolution of Knowledge

The knowledge about the city evolves and each new study can lead to revising assumptions or completing the knowledge about some objects that make up the city. For example, a simple photo exhumed from a newly found archive can narrow the range of a building's existence by demonstrating that it was built later than previously known. The evolution of knowledge also implies taking into account the evolution of the meaning of the vocabulary, in the representation of knowledge about the city. For example, a building that met accessibility standards for disabled people in the 1980s meets certain criteria that may not be sufficient in 2022, as the rules have evolved. The use of this type of information may therefore be different depending on when the building was declared compliant, but also on when this information appeared in the knowledge base. Moreover, the knowledge linked to the data itself can be used to intelligently publish this data. For example, when publishing data from the Lyon metropolitan area, the question arises as to whether published derived data will be identical to previously published data or whether a new version should be published. This raises the question of the reproducibility of a calculation, reproducibility linked not only to the initial data but also to the codes used to calculate the derived data. How can we ensure the evolution of knowledge, as well as the evolution of induced knowledge?

Sharing knowledge for everyone

Moreover, urban studies have often been built on mono-disciplinary approaches (in urban planning, geography, environmental science, sociology, economics, law, etc.). The resulting data and knowledge are therefore by construction heterogeneous [1,2,3,6]. In order to comply with regulations, local authorities have produced a very large mass of data that are quite uneven in terms of their documentation, quality, and capacity to represent the evolution of the territory [3,7,8]. In the need for transparency, it is also necessary to go further by delivering, in addition to the data, the algorithms that govern certain mechanisms of public life. An alignment of this knowledge [2,3,9] is necessary in order to aggregate this knowledge, but also to make it available to a large community. It is necessary to propose a diachronic approach allowing us to build knowledge in the long run. How can we allow a non-data scientist user to mobilize this heterogeneous knowledge?

Towards reproducible results for research and decision making

In addition, it is also important to integrate process and project knowledge into the scientific workflow in accordance with Linked Data principles. To this end, we propose to build an open and reproducible workflow to integrate [5,14], extract, question [6], reason, infer [11,12,13], and automatically generate new knowledge from multiple, autonomous, heterogeneous, and evolving urban data sources. Repeatability (obtaining the same results by repeating the exact same computations) and reproducibility (obtaining the same results by different means) are both key aspects of research. They allow us to confirm results or lead to a better understanding. The replicable approach will allow all stakeholders to not only access and interact with the data and knowledge produced, but also to verify and share the data, knowledge, code and this automated process to explain the reasoning behind their decisions. How to ensure knowledge related to the processes of production of data and algorithms to ease the reproducibility of results for both research and decision making?

Scientific challenges related to the project

In this thesis, several challenges are identified: ● Management of temporal graph data on knowledge (related to code, data, project, workflow)

● Ability to generate new information about derived knowledge; how to manage it in a context of knowledge versioning needs, in particular in a context of reproducibility of experiments.

● To propose new concepts of organization and interrogation of knowledge allowing it to scale.

Technical challenges related to the project ● Ability to propose a prototype based on the LIRIS Pagoda (https://projet.liris.cnrs.fr/pagoda/latest/) and UD-SV (https://github.com/VCityTeam/UD-SV/) platforms and to demonstrate the feasibility of the proposed approaches.

● Ability to propose the necessary tools for use by non-data scientists.

Application Interested persons should send a mail to Prof. Gilles Gesquière ([email protected]), Dr. Emmanuel Coquery (emmanuel.coquery@univ-lyon1. fr) and Dr. John Samuel ([email protected]) before September 1, 2022 with the following:

1. CV 2. Motivation letter 3. Academic transcripts

4. List of publications (if any) 5. 2 reference letters (or persons to contact)

As part of the Labex ASLAN and a project to help learn to write, the TWEAK team at LIRIS is looking for an engineer for the design and development of a tablet application in javascript.

Details of the offer can be found here: https://emploi.cnrs.fr/ Offres/CDD/UMR5205-FRELAF-002/Default.aspx

Starting date: November 1st, 2022 (flexible)

Application deadline: September 5th, 2022

Interviews (tentative): September 19th, 2022

Salary: ~2000€ gross/month (social security included)

Mission: research oriented (teaching possible but not mandatory)

*Keywords:*speech processing, fairness, bias, self-supervised learning,evaluation metrics

*CONTEXT*

The ANR project E-SSL (Efficient Self-Supervised Learning for Inclusive and Innovative Speech Technologies) will start on November 1st 2022. Self-supervised learning (SSL) has recently emerged as one of the most promising artificial intelligence (AI) methods as it becomes now feasible to take advantage of the colossal amounts of existing unlabeled data to significantly improve the performances of various speech processing tasks.

*PROJECT OBJECTIVES*

Speech technologies are widely used in our daily life and are expanding the scope of our action, with decision-making systems, including in critical areas such as health or legal aspects. In these societal applications, the question of the use of these tools raises the issue of the possible discrimination of people according to criteria for which societyrequires equal treatment, such as gender, origin, religion or disability... Recently, the machine learning community has been confronted with the need to work on the possible biases of algorithms, and many works have shown that the search for the best performance is not the only goal to pursue [1]. For instance, recent evaluations of ASR systems have shown that performances can vary according to the gender but these variations depend both on data used for learning and on models [2]. Therefore such systems are increasingly scrutinized for being biased while trustworthy speech technologies definitely represents a crucial expectation.

Both the question of bias and the concept of fairness have now become important aspects of AI, and we now have to find the right threshold between accuracy and the measure of fairness. Unfortunately, these notions of fairness and bias are challenging to define and their meanings can greatly differ [3].

The goals of this PhD position are threefold:

- First make a survey on the many definitions of robustness, fairness and bias with the aim of coming up with definitions and metrics fit for speech SSL models - Then gather speech datasets with high amount of well-described metadata - Setup an evaluation protocol for SSL models and analyzing the results.

*SKILLS*

- Master 2 in Natural Language Processing, Speech Processing, computer science or data science. - Good mastering of Python programming and deep learning framework. - Previous experience in bias in machine learning would be a plus - Very good communication skills in English - Good command of French would be a plus but is not mandatory

*SCIENTIFIC ENVIRONMENT*

The PhD position will be co-supervised by Alexandre Allauzen (Dauphine Université PSL, Paris) and Solange Rossato and François Portet (Université Grenoble Alpes). Joint meetings are planned on a regular basis and the student is expected to spend time in both places. Moreover, two other PhD positions are open in this project. The students, along with the partners will closely collaborate. For instance, specific SSL models along with evaluation criteria will be developed by the other PhD students. Moreover, the PhD student will collaborate with several team members involved in the project in particular the two other PhD candidates who will be recruited and the partners from LIA, LIG and Dauphine Université PSL, Paris. The means to carry out the PhD will be providedboth in terms of missions in France and abroad and in terms of equipment. The candidate will have access to the cluster of GPUs of both the LIG and Dauphine Université PSL. Furthermore, access to the National supercomputer Jean-Zay will enable to run large scale experiments.

*INSTRUCTIONS FOR APPLYING*

Applications must contain: CV + letter/message of motivation + master notes + be ready to provide letter(s) of recommendation; and be addressed to Alexandre Allauzen ([email protected]. eu), Solange Rossato ([email protected]) and François Portet ([email protected]). We celebrate diversity and are committed to creating an inclusive environment for all employees.

Starting date: October 01, 2022 (flexible)

Application deadline: September 5th, 2022

Interviews (tentative): September 12th, 2022

Salary: 1 975 € gross/month (social security included) Mission: research oriented (teaching possible but not mandatory)

Keywords: natural language processing, knowledge representation, cultural heritage, transfer learning, multilingualism

CONTEXT

The main challenge of the Patrimalp project is the development of an integrated and interdisciplinary Heritage Science, in order to ensure cultural Heritage sustainability, promotion and dissemination in contemporary society. The ambition is to produce the forms of intelligibility of a global and moving process which starts from the collection of the raw material, its transformation into a primitive object, different lives as a material (alterations, degradations, transformations ...) and as a symbol (relegation, disinterest, oblivion or rebirth, exaltation...) throughout history, and finally from its election as an object of historical and Heritage value and its “promotion” into a work of art. This research is applied to understand how inks and pigments have been conceived over several centuries, how it has been used in art work as well as how the handcrafting method has evolved and been disseminated over centuries and countries.

To make this study possible, the project will gather a large collection of textual material made up of alchemical works and collections of natural or artificial objects collected between the 16th and 18th centuries. To better understand the choice of colors for these "wonders", we want to reconstruct the recipes for making colored material in its context of thought, whether technical or symbolic. These recipes will constitute a new body of research for literary people and a new data-study case for building knowledge about color. This corpus indeed offers modes of representation inscribed in complex forms of writing and fiction whose modalities and frames of reference remain to be analyzed (accounts of technical, medical or physico-chemical experiments inscribed in fictional worlds or mythological, symbolic descriptions of artifacts, or materials collected in nature, mines). On the linguistic level, the inventory of this lexicon in different European and Eastern languages will lead to the formalization of the knowledge of these various skills over time and several cultures. This corpus will thus provide complex data on the material and symbolic origin of the ingredients of color, on its use, its names and its physical or symbolic perception: these data represent a challenge for computer researchers who will have to organize them in order to benefit curators, chemists or physicists, in ontologies representing the state of knowledge from the point of view of scholars over the ages. To systematically explore the corpus of these recipes, we will use NLP techniques to uncover the correlations between recipes, physical and chemical composition of objects and symbolic references. The final objective is to build a knowledge base (objects, components of objects, materials, colors, know-how, reference framework) each of the parts being able to reference a specific ontology (ontology of pigments, materials, colors...) to make it possible for researchers to observe the trajectory from the writing of color to its technical and artisan practice from this specific corpus.

PHD OBJECTIVES

The PhD project will focus on segmenting, extracting and representing recipes from a corpus of alchemical works from the 16th and 18th centuries to make them accessible to researchers in the humanities. This necessitates to :

- identify which excerpts of the text belong to a recipe; - supervise an annotation campaign to build an analysis and training corpus

- build NLP tools to extract automatically the list of elements (raw material, tools, quantity, units) and actions (verb, adverb, adjective) that made up the recipes; - analyze the dependencies between the elements of a recipe rules ; - Represent these rules in a formal knowledge representation.

The results of this processing will support : - The documentation of this unique set of text, by inserting the extracted elements to the document meta data to easy retrieval - The building a knowledge base of alchemical recipes

This PhD will need to address several challenges. One of them is to be able to process text composed of multiple non-modern languages (French, German, English, Latin, Greek) [Coavoux2022,Grobol2022] . One approach we will be to study how large multilingual pre-trained models [Delvin2019, Xue2020] can be leveraged and adapted for the task and how disparate collection of corpora of ancient texts can be used to fine-tune them. Another challenge will be the paucity of data for the downstream tasks (segmentation, parsing, Natural Language Understanding [Desot2022]) for this we will need to identify other related corpus (e.g. cooking) to address the problem in a multitask setting (such as NER and NLU) and transfer learning.

SKILLS

- Master 2 in Natural Language Processing, computer science or data science. - Good mastering of Python programming and deep learning frameworks.

- Previous experience in text classification, parsing, processing of several languages or text retrieval would be a plus - Very good communication skills in English and good command of French

SCIENTIFIC ENVIRONMENT

The thesis will be conducted within the STEAMER and GETALP teams of the LIG laboratory (http://steamer.imag.fr/ and https://lig-getalp.imag.fr/). The GETALP team has strong expertise and track record in Natural Language Processing, STEAMER team has strong expertise in Knowkledge representation and reasoning.The recruited person will be welcomed within the teams which offer a stimulating, multinational and pleasant working environment. The means to carry out the PhD will be provided both in terms of missions in France and abroad and in terms of equipment (personal computer, access to the LIG GPU servers). The PhD candidate will collaborate with the partners involved in the PATRIMALP project, in particular with Laurence Rivière from the LUHCIE lab (Laboratoire Universitaire Histoire Cultures Italie Europe) and Véronique Adam from the LITT&ARTS lab (Littératures et Arts).

INSTRUCTIONS FOR APPLYING

Applications must contain: CV + letter/message of motivation + master notes + be ready to provide letter(s) of recommendation; and be addressed to Danielle Ziebelin ([email protected]), François Portet ([email protected]) Maximin Coavoux ([email protected])

Starting date: November 1st, 2022 (flexible)

Application deadline: September 5th, 2022

Interviews (tentative): September 19th, 2022

Salary: ~2000€ gross/month (social security included)

Mission: research oriented (teaching possible but not mandatory)

*Keywords:* speech processing, natural language processing, self-supervised learning, knowledge informed learning, Robustness, fairness

*CONTEXT*

The ANR project E-SSL (Efficient Self-Supervised Learning for Inclusive and Innovative Speech Technologies) will start on November 1st 2022. Self-supervised learning (SSL) has recently emerged as one of the most promising artificial intelligence (AI) methods as it becomes now feasible to take advantage of the colossal amounts of existing unlabeled data to significantly improve the performances of various speech processing tasks.

*PROJECT OBJECTIVES*

Recent SSL models for speech such as HuBERT or wav2vec 2.0 have shown an impressive impact on downstream tasks performance. This is mainly due to their ability to benefit from a large amount of data at the cost of a tremendous carbon footprint rather than improving the efficiency of the learning. Another question related to SSL models is their unpredictable results once applied to realistic scenarios which exhibit their lack of robustness. Furthermore, as for any pre-trained models applied in society, it isimportant to be able to measure the bias of such models since they can augment social unfairness.

The goals of this PhD position are threefold: - to design new evaluation metrics for SSL of speech models ; - to develop knowledge-driven SSL algorithms ; - to propose methods for learning robust and unbiased representations.

SSL models are evaluated with downstream task-dependent metrics e.g., word error rate for speech recognition. This couple the evaluation of the universality of SSL representations to a potentially biased and costly fine-tuning that also hides the efficiencyinformation related to the pre-training cost. In practice, we will seek to measure the training efficiency as the ratio between the amount of data, computation and memory needed to observe a certain gain in terms of performance on a metric of interest i.e.,downstream dependent or not. The first step will be to document standard markers that can be used as robust measurements to assess these values robustly at training time. Potential candidates are, for instance, floating point operations for computational intensity, number of neural parameters coupled with precision for storage, online measurement of memory consumption for training and cumulative input sequence length for data.

Most state-of-the-art SSL models for speech rely onmasked prediction e.g. HuBERT and WavLM, or contrastive losses e.g. wav2vec 2.0. Such prevalence in the literature is mostly linked to the size, amount of data and computational resources injected by thecompany producing these models. In fact, vanilla masking approaches and contrastive losses may be identified as uninformed solutions as they do not benefit from in-domain expertise. For instance, it has been demonstrated that blindly masking frames in theinput signal i.e. HuBERT and WavLM results in much worse downstream performance than applying unsupervised phonetic boundaries [Yue2021] to generate informed masks. Recently some studies have demonstrated the superiority of an informed multitask learning strategy carefully selecting self-supervised pretext-tasks with respect to a set of downstream tasks, over the vanilla wav2vec 2.0 contrastive learning loss [Zaiem2022]. In this PhD project, our objective is: 1. continue to develop knowledge-driven SSL algorithms reaching higher efficiency ratios and results at the convergence, data consumption and downstream performance levels; and 2. scale these novel approaches to a point enabling the comparison with current state-of-the-art systems and therefore motivating a paradigm change in SSL for the wider speech community.

Despite remarkable performance on academic benchmarks, SSL powered technologies e.g. speech and speaker recognition, speech synthesis and many others may exhibit highly unpredictable results once applied to realistic scenarios. This can translate into a global accuracy drop due to a lack of robustness to adversarial acoustic conditions, or biased and discriminatory behaviors with respect to different pools of end users. Documenting and facilitating the control of such aspects prior to the deployment of SSL models into the real-life is necessary for the industrial market. To evaluate such aspects, within the project, we will create novel robustness regularization and debasing techniques along two axes: 1. debasing and regularizing speech representations at the SSL level; 2. debasing and regularizing downstream-adapted models (e.g. using a pre-trained model).

To ensure the creation of fair and robust SSL pre-trained models, we propose to act both at the optimization and data levels following some of our previous work on adversarial protected attribute disentanglement and the NLP literature on data sampling and augmentation [Noé2021]. Here, we wish to extend this technique to more complex SSL architectures and more realistic conditions by increasing the disentanglement complexity i.e. the sex attribute studied in [Noé2021] is particularly discriminatory. Then, and to benefit from the expert knowledge induced by the scope of the task of interest, we will build on a recent introduction of task-dependent counterfactual equal odds criteria [Sari2021] to minimize the downstream performance gap observed in between different individuals of certain protected attributes and to maximize the overall accuracy. Following this multi-objective optimization scheme, we will then inject further identified constraints as inspired by previous NLP work [Zhao2017]. Intuitively, constraints are injected so the predictions are calibrated towards a desired distribution i.e. unbiased.

*SKILLS*

- Master 2 in Natural Language Processing, Speech Processing, computer science or data science. - Good mastering of Python programming and deep learning framework.

- Previous in Self-Supervised Learning, acoustic modeling or ASR would be a plus

- Very good communication skills in English - Good command of French would be a plus but is not mandatory

*SCIENTIFIC ENVIRONMENT*

The thesis will be conducted within the Getalp teams of the LIG laboratory (https://lig-getalp.imag.fr/) and the LIA laboratory (https://lia.univ-avignon.fr/). The GETALP team and the LIA have a strong expertise and track record in Natural Language Processing and speech processing. The recruited person will be welcomed within the teams which offer a stimulating, multinational and pleasant working environment.

The means to carry out the PhD will be providedboth in terms of missions in France and abroad and in terms of equipment. The candidate will have access to the cluster of GPUs of both the LIG and LIA. Furthermore, access to the National supercomputer Jean-Zay will enable to run large scale experiments.

The PhD position will be co-supervised by Mickael Rouvier (LIA, Avignon) and Benjamin Lecouteux and François Portet (Université Grenoble Alpes). Joint meetings are planned on a regular basis and the student is expected to spend time in both places. Moreover, the PhD student will collaborate with several team members involved in the project in particular the two other PhD candidates who will be recruited and the partners from LIA, LIG and Dauphine Université PSL, Paris. Furthermore, the project will involve one of the founders of SpeechBrain, Titouan Parcollet with whom the candidate will interact closely.

*INSTRUCTIONS FOR APPLYING*

Applications must contain: CV + letter/message of motivation + master notes + be ready to provide letter(s) of recommendation; and be addressed to Mickael Rouvier ([email protected]), Benjamin Lecouteux([email protected]) and François Portet ([email protected]). We celebrate diversity and are committed to creating an inclusive environment for all employees.

- Duration: 3 years;

- Start of the contract: academic year 2022-2023;

- The thesis carried out within VITAL will be attached to MoDyCo, UMR 7114, /Doctoral School///Knowledge, Language, Modeling (ED 139) and at LAVUE, UMR 7218, Spaces, Time, Cultures doctoral school (ED 395).

*ELIGIBILITY CRITERIA* - Students who have not yet registered can apply in thesis and who have defended their Master 2 dissertation on the date of submission of applications.

- Candidates may be students who have continued their university course in France or abroad.

*SELECTION CRITERIA * The selection criteria for this doctoral contract will be the following:

- the scientific quality of the dossier (clarity of the problem, methodology and methods of implementation);

- the quality of the candidate's career;

- the match between the candidate, his/her thesis project and the profile of the “VITAL” doctoral contract;

- knowledge and mastery of learning tools automatic (machine learning, deep learning) and annotation manual;

- the ability to work in a group with specialists from other disciplines;

- interest in the field of town planning and the development of space in general as well as for multimodal data.

*SELECTION METHODS* The judging panel will be made up of the following people:

- 1 external member mandated by the Labex Past in the present;

- 1 member representing INA;

- The scientific manager of the Labex Past in the present or the person authorized to represent him;

- 1 member mandated by the ComUE UPL Doctoral College (one of the two co-directors);

- a representative of the ComUE presidency team Université Paris Lumières, which chairs the jury.

The auditions are planned in person at the headquarters of the University of Paris Lights, 75013 PARIS. This method may be modified depending on ministerial directives related to the COVID19 pandemic.

*THE APPLICATION FILE*

Candidates who wish to apply for this doctoral contract must provide a file consisting of the following documents:

1. The completed application file (including the thesis project: 10,000 characters maximum. Font Time, size 12, bibliography summary included. Please note: for the sake of equality, the pages will be removed from the file);

2. An academic curriculum vitae (2 pages maximum);

3. A cover letter (3,000 characters maximum);

4. Master's transcript;

5. The master's degree or, failing that, the defense certificate;

6. The Master's thesis (in .pdf);

7. The reasoned opinion of the thesis supervisor anticipated;

8. The reasoned opinion of the management of the host research unit.

The application file will be sent in electronic form in the format .pdf (a *single file* bearing the name of the candidate) on the address following: pasp.allocations22[at] passes-present.eu

In the "subject" of this email, please specify expressly as according to the following example: Candidature-DOC-VITAL

An acknowledgment of receipt will be sent thereafter.

*CONTACT*

For any additional information about this call for applications (outside definition of the scientific project of the thesis), write to the following address: pasp.allowances22[at] passes-present.eu

*To contact the scientific team of the VITAL* project, please contact:

Iris ESHKOL-TARAVELLA : ieshkolt[at] parisnanterre.fr

Olivier RATOUIS: oratouis[at] parisnanterre.fr

*To contact the INA: * contact Géraldine POELS: gpoels[at] ina.fr

*The Labex VITAL project The past in the present*

The VITAL project brings together researchers in an original and innovative way automatic language processing (TAL), architects and town planners, in a partnership research project with the National Archives and in collaboration with other institutions, including INA. This project aims à to understand how the speeches and projects architectural and urban approach space as a dynamic and study how this dynamic is expressed. We make the assumption that, just as much as by spatial alternatives, the design and the organization of cities are structured by relationships to time contrasting based on differentiated perceptions and conceptions of urban change, and that it is essential to study them in order to understand how societies project themselves into space.

*Context of the research project*

Urban modernity is characterized by the conduct of interventions in the city, through voluntary and rational types of action linked to the discipline of town planning which appeared with the 20th century. Developing the modern city means making its space rational for society that inhabits it. It thus appears that urban modernity achieves a far-reaching change in societies Western cultures, which establishes in a more specific way a new framework life, breaking with previous periods. In the years that following the Second World War, this project takes on a new magnitude.

If these changes have been studied, this is less the case of their reception and appropriation by city dwellers and the various actors in society. Although decisive, this aspect calls for proof of a renewed approach to identifying and measuring them. He therefore becomes interesting to identify indicators of places and of moments, of events, in which and through which are manifested these deep dynamics.

We will focus here on public forms of radio and television expression, which carry both the discourse (understood in the broad sense of oral, written and visual) of planners and that of the public, in this case the inhabitants and townspeople of the modernized city. The partnership offered by the National Institute of audiovisual (INA) thus offers an exceptional opportunity.

To do this, the project plans to identify and analyze the emotions which come under these discourses by using audiovisual sources, but the candidate can suggest other approaches and tools that seem judicious to him.

*Project partners*

The call for applications is based on an important collaboration and original between different partners. The doctoral student will benefit from a privileged welcome within the INA Lab which will be inaugurated in September 2022. This Lab will offer a range of services and tools for the automated exploitation and analysis of corpora from from the INA collections and will be able to offer the doctoral student scientific and technical support.

*Supervision of the **doctoral student*

The doctoral student will be co-supervised by the professors Iris ESHKOL-TARAVELLA (ED 139 “Knowledge, Language, Modeling”) and Olivier RATOUIS (ED 395 - Spaces, Times, Cultures).

*Research entrusted to the doctoral student*

In partnership with INA, the research aims to develop a tool automatic annotation of emotions found in the corpus multimodal projects dealing with land use planning and urban planning in using the tools and techniques of Automatic Language Processing (TAL).

The observation of the emotions detected will make it possible to study the perception of territorial development from a new angle. The resources constituted will be used by the INA, and shared and made available to the scientific community and the general public.

*Doctoral student profile* The disciplinary profile of the doctoral student is open to holders of an M2 TAL diploma. An interest in the fields of urban planning and Artificial Intelligence, as well as the experience work with multimodal corpora will be appreciated.

A good command of the French language is expected.

The context : The National Audiovisual Institute is looking for an Engineer of research for the INA Lab on permanent contracts, attached to the Research Department of the Data & Technology Department. Within the Directorate of Heritage, the INA thèque is developing a specific service offer research work centered on the exploitation of data sets massive from the INA collections.

Gateway between Management Data & Technologies and the Heritage Department, the INA Lab newly created will be both a research facilitation structure and incubator of innovative projects centered on the exploitation of data via automated search tools and processing, analysis or visualization.

EPIC, created in 1975, INA is responsible for conserving, promote and transmit the French audiovisual heritage. First digital archive center in the world with more than 22 million hours of television and radio, to which are added each year 1 million hours for legal deposit, INA has approximately 1,000 employees.

Assignment : Organize the support and accompaniment of projects and works of research which exploit the Radio, TV and web funds of INA and require the implementation of technical solutions, the development or deployment of tools and systems adapted to centered approaches on data analysis and mining technologies.

Activities / tasks: - Coordinate and organize the operational support of the works and research projects that apply methods and tools automated analysis to large corpora of data from INA collections

- Develop a support protocol for research projects data-oriented or automatic processing - Manage, from generic solutions and excavation tools and analyzes developed at INA, the functional analyzes of needs of researchers and draft specifications techniques in collaboration with researchers and bearers of projects (data processing and enhancement)

- Deploy and maintain in operational condition tools and data enhancement services by favoring an approach "open source" and free software in connection with the teams INA techniques and R&D

- Design and implement a management strategy software and technical resources in coordination with the team INA technology library - Organize a scientific and technological watch on the methods and tools adapted to new scientific uses of analysis and data mining

- Play an advisory role with users, researchers and students - Participate in the scientific animation in connection with the projects accompanied (Datasprints, MasterClass, Seminars, Training ...)

Specific activities - Support and coordinate scientific research projects - Design and deploy devices and services for digital processing, enhancement and mediation of data - Know how to represent and promote research and work accompanied at the INA thèque - Support the evolution of professional practices and organize training INA employees in tools and procedures

Profile / Skills: - Justify a diploma (BAC +5) in engineering or a Master's degree in documentary computing or digital humanities ideally doubled a higher education in another discipline, in Sciences human and social and 3 years of experience in support or management of scientific work oriented searching and processing of data, attested by participation in projects and possibly publications.

- Mastery of the theoretical and practical issues of the Humanities knowledge and in-depth knowledge of technologies and systems documentary information - Knowledge of institutional and scientific networks in the fields of human and social sciences and digital humanities - Mastery of quantitative analysis methods - Good knowledge of data processing tools and technologies data (Python programming language, analysis software R statistics) - Knowledge of data formats (XML, Json) - Knowledge of the principles of building web interfaces (HTML, CSS...) - Knowledge of REST APIs

Please send your CV and cover letter to: [email protected] , [email protected] , [email protected]

Find this offer online: https://www.ina.fr/nous-rejoindre/offres-emploi/ingenieurvde-recherche

post-doctorate in TAL at LISN Identification of gendered expressions by representations vectors on a corpus of speech transcription in the media

Type of contract: CDD 1 year https://emploi.cnrs.fr/Offres/CDD/UMR9015-CYRGRO-001/Default.aspx

*1. Tasks* The GEM (Gender Equality Monitor) project aims to analyze the interactions between women and men in the media (radio and television), and more particularly the differences in representations depending on whether the person speaking is a woman or a man, according to his role (anonymous, journalist, politician, etc.), and according to the themes.

In this interdisciplinary project, the partners (including LISN) are responsible for implementing the descriptors that will allow partners in the human sciences and to quantify and qualify differences in representation. https://anr.fr/Projet-ANR-19-CE38-0012

*2. Activities* The recruited person (M/F) will be in charge of developing unsupervised automatic language processing (TAL) techniques or semi-supervised applied to corpora of transcriptions automatic speech, to identify "gendered expressions" such as references to cultural stereotypes based on the gender, traditional named entities or any reference to life privacy, age, physique, sexuality, skills, etc. Secondarily, the analysis of biases in language models can also be driven.

The corpora are made available by the project leader (Institut National de l'Audiovisuel) and consist of: morning radio and television logs from the GMMP corpus (Global Monitoring Media Project), French radio programs (cooking shows, economic, sporting, and free-to-air) for the study of incivilities (interruptions, insults, etc.), and reality TV shows (Loft Story 2001, The Marseillais in Dubai 2021). No annotation is available around gendered expressions. The recruited person must therefore favor unsupervised or semi-supervised methods.

This work will be co-supervised by Ms. Sahar Ghannay (MCF in computer science at Paris Saclay University) and Mr. Cyril Grouin (IR in computer science at CNRS). The contract will be funded by the National Research Agency (ANR GEM 2019) led by David Doukhan (National Institute of Audiovisual).

*3. Skills* - very good command of French - automatic language and speech processing; a training specific in this discipline is a plus - experience of lexical embeddings and neural networks

*4. Work context* The Interdisciplinary Laboratory of Digital Sciences (LISN) is a unit installed on the Saclay plateau and created in 2021 from the merger of LIMSI and LRI laboratories. Research carried out at LISN cover a broad scientific spectrum and are recognized the international.

The laboratory includes more than 380 members divided into 16 teams of research and 6 support services and support. The premises are entirely in a restricted regime zone (ZRR).

The recruited person will work within the ILES team, in connection closely with the researchers from the ILES and TLP teams involved in the project, within the Language Sciences and Technologies department (STL).

*5. Constraints and risks* Travel possible in Ile-de-France for work meetings punctual

National and international travel in conference in case of article to present computer work

context Collaboration between a LIMOS / CNRS laboratory from Clermont Auvergne University and a young innovative company Jeolis Solutions, both located in Clermont-Ferrand, in the field of e-Health and more specifically Therapeutic Patient Education (ETP[ 1] ). Environmental constraints and the health crisis such as that of Covid-19 plead for an absolute need to quickly develop intelligent digitalization in order to achieve a motivating, fun and personalized e-ETP.

Tasks In the perspectives of the article [1], it was stated the possibility of improving the logical expressiveness of the program by using other approaches such as ASP (Answer Set Programming) in order to limit the use of imperative code.

The missions of the recruited person will be the implementation of tools allowing to:

Combining the best of the OWL2 and ASP worlds OWL2 for the static aspect of knowledge (TBox and ABox + consistency check by classification of individuals), like what is offered in the Protégé editor. In addition OWL2 allows inheritance between classes and relations.

ASP for the dynamic aspect of knowledge and the relevance of logical rules with negation(s), constraints and non-monotonic

The theoretical aspect of this hybridization has been proposed in Hexlite [2] but we believe that an implementation in all Python would facilitate adoption.

Establish a DMN to ASP transcription Writing an ASP program is not yet within the reach of business experts in the industrial world, which does not facilitate the adoption of ASP despite powerful solvers. However, there is the DMN standard (Decision Model and Notation) of the OMG consortium which allows manufacturers to write decision systems from simple Excel-type tables. It would be interesting to have a tool allowing the transition from DMN to ASP, like what was proposed recently in cDMN [3].

The interest would be to allow expert psychologists and pedagogues to write, improve and test progression strategies more easily, more quickly than at present, without going through an expert engineer specialized in knowledge engineering.

Work context The recruited person will join the Data, Services, Intelligence theme of LIMOS but will be made available to the Jeolis Solutions R&D team for 80% of their time (application of the France Relance plan – preservation of R&D employment in companies) .

The CDD is for a period of 19 months, and must start before the end of September 2022, adaptable according to the availability of the person selected.

Expected skills · Holder of a master's degree in computer science (obtained in 2019, 2020, 2021 or 2022) with a solid experience in Python development · Knowledge/experience of semantic web standards (OWL2, SWRL…) and also in logic programming (PROLOG, Answer Set Programming…)

· Knowledge in Knowledge Engineering · Experience with the Git code manager tool and Agile methodology · Respect for deadlines, rigor, organization, ability to work in multidisciplinary teams

How to apply

Candidates are invited to submit a CV, a cover letter and possibly reference letters, in PDF format to [email protected] , to [email protected] and to [email protected] .

Applications will be processed on an ongoing basis.

The Ph.D. positions are full-time and for maximum 3.5 years. The salary is about 40K€ per year. The research, publications, and travelling cost will be covered separately by the BMW Group.

The Internships and Master thesis positions are for maximum one year.

Topics of Ph.D. positions:

​Topic 1: Robust multi-modal perception and learning with performance guarantees in robotics and HMI

​Topic 2: Visuo-Tactile counterfactual-based deep active learning and decision making in robotics and intelligent interactive systems

Topic 3: Machine learning-based optimal charging policies to improve vehicle's battery life

Topics of Internship positions:

Embedded Software Programmer Control design and sensor/actuator for intelligent systems AI/ML on-Chip optimization and Computing Robotics Grasp and Manipulation SLAM and Navigation

Eligibility MSc degree in Electrical Engineering, Computer Science Mathematics, Systems and Control, or a related field from only top international universities. Strong theoretical or mathematical background, and a strong interest in AI, Robotics and Machine Learning.

Advanced programming skills in ROS, C++ and Python are mandatory. The ideal candidate should be familiar with Issac Nvidia, and various machine learning and robotics libraries.

Having at least one publication in top IEEE conferences such as ICRA, IROS, ICML, CVPR, NeurIPS or top journals such as T-RO, AURO, RAL, science robotics, and etc. Efficient communication skills in English.

How to Apply:

Required documents

CV with photo One-page cover letter (clearly indicating available start date as well as relevant qualifications, experience and motivation) University certificates and transcripts (both BSc and MSc degrees) Contact details of up to three referees List of publications

How to send your documents

All documents should be in English and in one single pdf saved with your first name and family name. ( Otherwise, it will be deleted)

Please send your application to: [email protected] by 15th August 2022. Top application will be invited for an oral presentation and interview. The project will start on October 1st, 2022.

For any questions regarding topic, positions, etc,.. please contact [email protected] or [email protected]

BMW Group Dr.-Ing. Mohsen Kaboli Asst. Professor of AI and Robotics

Head of the BMW AI Robotics Lab and RoboTac Team Head of the AI Robotic and Tactile Intelligence Group at Donders Institute for Brain and Cognition, Radboud University

Department of Research, New Technologies, Innovations AI Robotics Center of Excellence Parkring 19 85748 Munich, Germany

In this project, we will investigate abstract principles of two-stream vision systems in the context of deep neural networks and sensor fusion. The aim is to develop new real-world applications of machine vision using two-stream video or laser sensors. The project will be conducted in collaboration with industry partner 4Tel Pty Ltd in Newcastle, Australia. The PhD candidate will be part of an interdisciplinary team of researchers from computer science and industry.

Please find further details at: https://www.newcastle.edu.au/study/research/phd-scholarships/phd-scholarships/neural-principles-of-dual-sensor-vision-systems-in-machine-vision-applications

Closing Date: 1 October 2022

Contact:

Professor Stephan Chalup School of Information and Physical Sciences The University of Newcastle, Callaghan NSW 2308, Australia Email: [email protected]

Inria Rennes, France

1 Membership Inference

One of the wonders of machine learning is that it turns any kind of data into mathematical equations.

Once you train a machine learning model on training examples—whether it’s on images, audio, raw text, or tabular data—what you get is a set of numerical parameters.

In most cases, the model no longer needs the training dataset and uses the tuned parameters to map new and unseen examples to categories or value predictions.

You can then discard the training data and publish the model on GitHub or run it on your own servers without worrying about storing or distributing sensitive information contained in the training dataset.

Nevertheless, a type of privacy-leak oriented attack against ML systems, namely membership inference, makes it possible to detect whether a given data instance was used to train a machine learning model.

In many cases, the attackers can stage membership inference attacks without having access to the machine learning model’s parameters. They just query the model and observe its output (soft decision scores or hard predicted labels).

Membership inference can cause severe security and privacy concerns in cases where the target model has been trained with sensitive information. For example, identifying that a certain patient’s clinical record was used to train a automatic diagnosis model reveals that the patient’s identity and relevant personal information.

Moreover, such privacy risk might lead commercial companies who wish to leverage machine learning-as-a-service to violate privacy regulations. [VBE18] argues that membership inference attacks on machine learning models increase greatly the vulnerability of machine learning service providers on privacy leaks.

They may face further legal issues related to privacy information breaching in their business practices due to GDPR (General Data Protection Regulation).

2 Thesis

In this thesis, our plan is to first implement and benchmark typical membership inference attacks proposed in the literature [LZ21, SDS+19, SSSS17, CCTCP21, CCN+22].

We need to carefully outline the impact of crucial parameters such as the hardness of the classification task (dimension of the inputs, number of classes), the size (depth, number of parameters), the training procedure (data augmentation), and the potential overfitting of the target model. This also includes the working assumptions about the attacker’s knowledge on the training data and his computation power. Indeed, some attacks rely on unrealistic assumptions. Designing more tractable attacks is key in order to clearly define when membership attacks are a real threat in practice.

In differential privacy [ACG+16, NST+21], a common defense is to randomize the procedure by adding noise either on the inputs (the training data set), the training procedure of the model, or the outputs (the trained model’s parameters).

This idea witnesses several implementations in modern machine learning like randomness in label smoothing, data augmentation, or penalization.

The study focuses on evaluating the multiple trade-off between the loss of classification performance, the prevention of overfitting, and the gain of robustness against membership inference attacks but also against adversarial attacks [SSM19].

Beyond inferring the membership of a given instance, we will also study the feasibility of attribute inference attack targeting to reversely estimate the attributes of training data, which is an extension to membership inference.

The candidate for this thesis is expected to have accomplished courses on Machine Learning and/or have experience of implementing Machine Learning algorithms using Python for practical data mining problems. Especially, expertise in using Pytorch will be required in the project.

Theoretical developments are also expected based on statistics and theory of machine learning and approximation.

The thesis takes place within INRIA Rennes, campus universitaire de Beaulieu, Rennes, France.

In the context of the research project “MultimEdia Entity Representation and Question Answering Tasks” (MEERQAT – ANR 2020-2024), a postdoctoral position is proposed for highly motivated candidates interested in computer vision and multimedia understanding.

https://www.meerqat.fr/wp-content/uploads/2021/06/meerqat_postdoc_cea_inria.pdf

under the aegis of the Ministry of Economy, Beyond5G project focuses on securing the network slice in 5G/6G environments, among other topics. In this project, we would like to strengthen our staff by hiring a doctor (Ph.D)

in computer science to support us in evaluating the security of 5G/6G slices using digital twins.

To apply, please follow the link below: https://institutminestelecom.recruitee.com/l/en/o/postdoctorat-in-digital-twin-for-quantifying-the-security-of-5g6g-slices

NOTE : the application period has been extended to end of August 2022. Applicants are not required to hold French or European citizenship.

*Job: Infolinguist M/F on permanent contracts, Dassault Systèmes Paris*

*Imagine tomorrow…*

You join Dassault Systèmes, and the team in charge of the software Proxem Studio as infolinguist.

As part of each project, and under the supervision of a head of project, the infolinguist designs, develops, tests and maintains the linguistic resources necessary for the successful completion of the project (quality, time and charge). He/she participates, with the project manager, to the development of certain deliverables (mainly plan of classification and platform configured for the project).

*Your missions:*

- Participate in project scoping meetings in order to take knowledge of the context and the objectives of the project, the data, the deliverables to be produced and the schedule.

- Integrate the corpus of customer data within the platform, with the help of a third person if necessary.

- Carry out an exploratory study of the client corpus. - Create and extend the linguistic resources necessary for the successful completion of projects by following the directives of the head of project.

- Prepare and participate in workshops with the client under the management of the project manager.

- Participate in follow-up points set by the project manager, alert in case of deviation from the project plan. - Ensure, throughout the project, the quality of resources linguistics by controlling their precision and coverage through assessments and error digs.

- Ensure that the semantic analysis and configuration of the platform for which the project manager has entrusted him with the responsibility allow an exploitation of the results in accordance with the needs and customer expectations

- Document the settings implemented throughout the project. - Ensure the maintenance of the linguistic resources created and manipulated.

*Your strengths for success:*

- You hold a Bac +5 in linguistic engineering - You already have a first experience in the field, internships and/or learning included

- You know the rules of the technical art and organizational skills in linguistic engineering (TAL, linguistics formal, formal languages…)

- You are able to carry out corpus studies (exploration of data, understanding of domain vocabulary, frequency expressions, complexity of the analyzed language, etc.) - You know how to set up semantic analyzes adapted to the project needs

- You know the fundamental concepts of the evaluation of the quality (precision, recall, etc.) - You master linguistic and semantic analysis tools (and you will be trained in Proxem Studio) - You master office software (text editor and spreadsheet)

*Soft Skills*

You are recognized for your adaptability, your flexibility, and your ability to interact, whether orally or in writing. You have excellent analytical and synthesis skills, and proof of creativity and sense of innovation. You appreciate teamwork, and have a customer-oriented mindset.

*To apply*: follow the "Apply" link on the page https://www.3ds.com/fr/careers/jobs/nlp-industry-linguist-scientist-specialist-527884

*Call for applications*

*Research contract* *within the VITAL project (City, Temporalities, Language)

Labex The past in the present Session 2022/2023*

*The past in the present labex is recruiting a beneficiary postdoctoral researcher starting January 2, 2023 and for 8 months in the under a postdoctoral contract assigned to the VITAL project.*

*Key terms:* urban planning, architecture, NLP, spatial dynamics, urban project, urban change, discourse of city specialists

*GOALS*

The VITAL project (Cities, Temporalities, Language) associates in a way original and innovative researchers in natural language processing (TAL), architects and urban planners, in a research project partnership with the National Archives and in collaboration with other institutions. This project aims to understand how speeches and architectural and urban projects address space as a dynamic and to study how this dynamic is expressed.

The interdisciplinary approach adopted here mobilizes the treatment automatic languages ​​(TAL) in order to identify and analyze elements of discourse reflecting the notion of urban dynamics and its perception by specialists in the field of urban planning in planning projects and initiatives. This is to test the feasibility and reliability of an automatic linguistic analysis of expressions referring to the dynamics in the field of urban planning as well as to propose a modeling of the phenomena sought and a typology of markers characterized by associated properties.

The focus will be on three areas:

- modeling of the information sought; - development of a tool for the automatic detection of this information ; - extraction and visualization of annotated information.

*ADMISSION REQUIREMENTS*

Candidates must hold a doctoral thesis in automatic processing of natural language or in tooled linguistics and show an interest in the field of urban planning. Autonomy in coding in python is essential, as well as basics in machine learning.

*CALENDAR*

- Deadline for submitting applications: November 14, 2022 at noon (Paris time) - Selection of applications: between November 16 and November 30 2022 - Announcement of results: December 1, 2022 - Starting date: January 2, 2023

Detailed job description and application procedures on the website of the labex:

http://passes-present.eu/fr/appel-candidatures-contrat-de-recherche-projet-vital-ville-temporalites-langage-44610