18 Postdoctoral Fellowship at Inria, France

Atlantis is  a joint project-team  between Inria and  the Jean-Alexandre Dieudonné Mathematics Laboratory at  Université Côte d’Azur. The team  gathers applied mathematicians and  computational scientists who are collaboratively undertaking  research activities aiming at the design, analysis, development and  application of advanced numerical methods for solving systems of  partial differential equations (PDEs) modelling nanoscale light-matter interaction problems. In this context, the team is  developing  the   DIOGENeS  [https://diogenes.inria.fr/]  software suite,  which  implements  several Discontinuous  Galerkin  (DG)  type methods tailored to the systems  of time- and frequency-domain Maxwell equations  possibly coupled  to  differential  equations modeling  the behaviour of propagation  media at optical frequencies. DIOGENeS also includes a component dedicated to the optimization of geometrical characteristics of nanostructures driven by some performance objective in the contex of inverse design strategies of nanophotonic setups.   DIOGENeS is a unique  numerical   framework  leveraging   the  capabilities   of  DG techniques  for  the simulation  of  multiscale  problems relevant  to nanophotonics and nanoplasmonics.

This project is based on a collaboration between Scool and a medical team at Inserm/CHU de Lille that have been active for 5 years now.

The scientific goal of this collaboration is to investigate the exploitation of data to improve patient follow-up after surgery. We model this problem as a sequential decision making under uncertainty problem.

This collaboration has already produced interesting results, in the form of a website, presentations in conference, and publications in top machine learning conferences, and top medicine journals.

This collaboration involves about 10 people: data engineer, PhD student, post-doc, researchers, professors.

This post-doctoral position will be supported by the  HE Trumpet project, the HE Flute project and/or the PEPR IA Redeem project.      While this position will be in the MAGNET team in Lille, we will collaborate with the several European project partners.

While AI techniques are becoming ever more powerful, there is a growing concern about potential risks and abuses. As a result, there has been an increasing interest in research directions such as privacy-preserving machine learning, explainable machine learning, fairness and data protection legislation.
Privacy-preserving machine learning aims at learning (and publishing or applying) a model from data while the data is not revealed. Notions such as (local) differential privacy and its generalizations allow to bound the amount of information revealed.

The research project will explore quantum protocols based on the concept of quantum nonlocality and quantum networks (see arXiv:2104.10700). A non-exhaustive list of potential projects is:

• Methods for characterizing quantum correlations beyond the Bell scenario:
  • mathematical foundation of these methods (C* algebras, noncommutative polynomial optimization): see e.g. arXiv:2210.09065, arXiv:2212.11299, arXiv:2301.12513
  • improve/find new algorithms for characterizing these correlations
  • numerical developpement of these algorithms, see e.g. arXiv:2211.04483
• Understanding the foundational implications of quantum correlations in networks, see e.g. arXiv:2101.10873 and arXiv:2105.09381
• Develop the applications of network nonlocality to certification protocols, such as
  • randomness generation: arXiv:2209.09921 
  • self testing of measurements and states: arXiv:1807.04956, arXiv:2201.05032
• Adapt existing protocols for their experimental implementation
• Develop practical benchmarks of the concept of ‘Genuine Multipartite Nonlocality’ introduced in arXiv:2105.09381
• Develop SDP relaxations for condensed matter problems, see e.g. arXiv:2212.03014, arXiv:2310.05844, arXiv:2311.18707, arXiv:2311.18706 
• Explore the limits of quantum distributed computing, see e.g. arXiv:1810.10838, arXiv:0903.113

The appointment is in the broad intersection of information theory, communication theory, and game theory. As part of the national programme France 2030, funded by the French Government, the objective of this position is to study the fundamental limits on the reliability and throughput of decentralized communications systems. The underlying assumption is that independent receivers might broadcast a signal that can be observed by all receivers and thus, used to enhance reliability, throughput, or a convex combination of both. More importantly, transmitters are assumed to use a final constellation and a limited number of channel uses (finite block-length regime). 

Given an AI model’s placement in an IDN, we will study inference delivery strategies to be implemented at each node in this task. For example, a simple inference delivery strategy is to provide the inference from the local AI model if this seems to be accurate enough or to forward the input to a more accurate model at a different node if the inference quality  improvement (e.g., in terms of accuracy) compensates for the additional delay or resource consumption. Besides this serve-locally-or-forward policy, we will investigate more complex inference delivery strategies, which may allow inferences from models at different clients to be combined. To this purpose, we will rely on ensemble learning approaches [MS22] like bagging [Bre96] or boosting [Sch99], adapting them to IDN distinct characteristics. For example, in an IDN, models may or may not be trained jointly, may be trained on different datasets, and have different architectures, ruling out some ensemble learning techniques. Moreover, queries to remote models incur a cost, which leads to prefer ensemble learning techniques that do not require joint evaluation of all available models.

The goal of this postdoc project is to investigate how to optimize resource and task allocation when deploying data streaming processing applications in the Fog. In particular, we want to investigate new optimization metrics and objectives when deploying streaming processing applications in the Fog, including latency, throughput, and maximum sustainable throughput. Accordingly, we will develop a new scheduling framework that relies, among others, on Machine Learning/Deep Learning models to decide on resource allocation and operator placement at runtime (based on the collected data and given the cost model of redeployment and process migration). The proposed framework will be integrated in one of state-of the art data stream engines such as Flick [6], Storm [7] or Spark [8] and evaluated at large-scale using syntactic applications and real-world stream data application.

Ce poste concerne le groupe de recherche Parole&Audio d’Inria Défense&Sécurité. Il est composé d’une dizaine de personnes, réparties majoritairement sur les sites Inria de Paris et Grenoble. Ce groupe développe des recherches principalement en détection du locuteur, de la langue parlée et de deepfakes audio. Il s’intéresse également à la transcription de la parole ainsi qu’à l’extraction d’informations de plus haut niveau, comme les entités nommées.

La spécificité du groupe de recherche Parole&Audio est de s’intéresser à des problématiques relevant de besoins exprimés par différents services de l’état autour des questions de Défense et de Sécurité, et de placer les potentiels opérateurs au cœur des travaux qui sont menés, que cela soit dans le cadre de travaux de recherche appliquée ou plus fondamentaux.

Ce poste concerne le groupe de recherche Parole&Audio d’Inria Défense&Sécurité. Il est composé d’une dizaine de personnes, réparties majoritairement sur les sites Inria de Paris et Grenoble. Ce groupe développe des recherches principalement en détection du locuteur, de la langue parlée et de deepfakes audio. Il s’intéresse également à la transcription de la parole ainsi qu’à l’extraction d’informations de plus haut niveau, comme les entités nommées.

La spécificité du groupe de recherche Parole&Audio est de s’intéresser à des problématiques relevant de besoins exprimés par différents services de l’état autour des questions de Défense et de Sécurité, et de placer les potentiels opérateurs au cœur des travaux qui sont menés, que cela soit dans le cadre de travaux de recherche appliquée ou plus fondamentaux.

Le poste proposé s’insère dans un projet financé par la DGA dans le cadre de la convention DGA-Inria sur l’Intelligence Automatique. Le projet vise à développer des solutions en caractérisation du locuteur et de la langue parlée parlée au sein de grandes collections de documents audio ou video. L’objectif est d’obtenir une solution qui soit performante, avec un haut niveau d’explicabilité et frugale. La frugalité s’exprime ici sur deux dimensions. La première concerne la consommation de ressources de calcul/mémoire pour l’apprentissage, l’adaptation des modèles et l’inférence. Ce point a pris une importance particulière avec l’arrivée de modèles dits « backbone », pré-entrainés et de très grande taille. La deuxième dimension de frugalité concerne l’incrémentalité des solutions proposées, soit à minimiser les coûts lorsqu’un nouveau jeu de données est inséré dans une très grande base de données déjà traitée.

Inria, the French National Institute for Computer Science and Applied Mathematics, promotes “scientific excellence for technology transfer and society”. Graduates from the world’s top universities, Inria’s 2,700 employees rise to the challenges of digital sciences. With its open, agile model, Inria can explore original approaches with its partners in industry and academia and provide an efficient response to the multidisciplinary and application challenges of digital transformation. Inria is the source of many innovations that add value and create jobs.

As part of the “CMN Numérique” program, which aims to standardize the production of 3D models and reuse them in various contexts (mediation, maintenance, research, etc.), the National Monuments Center (CMN) has produced a collaborative 3D experience, including an ultra-realistic reproduction of the Château d’Azay-le-Rideau, in the Centre Val de Loire region of France. Developed in Unreal Engine 5, it includes a 3D model of the château’s exterior. Today, the park has been reconstructed including the water mirror, the staircase and four Renaissance rooms. The next step is to virtually furnish these rooms using with around 10 pieces of furniture (beds, chests, chests of drawers).

This postdoctoral position is part of the 3DS-INRIA MediTwin project, aimed at developing a preoperative surgical planning tool to support medical teams in managing single ventricle patients.

Once the surgical procedure is determined by the 3DS preoperative planning tool, the goal is to provide a simplified digital twin (0D model or other reduced model) that incorporates morphological changes from the planning tool as well as patient-specific functional data (e.g., from pulse wave Doppler ultrasound). This 0D digital twin can be calibrated using either a high-fidelity 3D digital twin or direct patient data, ultimately enabling postoperative outcome prediction and improved therapeutic decision-making.

The postdoctoral researcher will join the multidisciplinary and international Simbiotx Inria research team (https://team.inria.fr/simbiotx/) and will closely collaborate with cardiopediatric teams at 3DS, the Inria Commedia team for high-fidelity models, and the Necker-Enfants Malades Hospital.

Daniel Stilck Franca and Mischa Woods are looking for candidates working in one or more of the following:

• Characterization and benchmarking of quantum devices.
• Dissipative preparation of quantum many-body states.
• Characterization of dynamical quantum advantages.

The focus of the research will be on quantum simulators/analog devices. Candidates with a diverse background in computer science, mathematics, or physics are welcome to apply.

Dans le cadre du projet mentionné ci-dessus, l’objectif de ce poste est de proposer et développer un algorithme d’apprentissage par renforcement (AR) multi-agents pour la navigation de robots sociaux.

Le travail consistera tout d’abord à adapter une approche récente d’apprentissage pour la navigation sociale multi-robots proposée dans l’équipe. Cette approche [1] s’inspire de récents travaux en AR multi-agents basés sur des graphes d’attention [3-4] pour étendre des modèles de navigation sociale mono-robot de l’état de l’art [2].

Le premier objectif sera d’améliorer et d’étendre cet algorithme pour permettre à une flotte de robots de naviguer dans un environnement présentant une complexité proche de celle de l’environnement réel visé dans le projet Solar-Nav (un hôpital). Il faudra en particulier intégrer la gestion des obstacles statiques et/ou dynamiques non humains. Cette intégration nécessitera de rajouter des observations compatibles avec la perception robotique (e.g. Lidar, Point Cloud Library).

 

(18) Postdoctoral Fellowship Position

Postdoc Fellowship Position summary/title: Post-Doctoral Research Visit F/M Data-driven metamaterial identification: from wavefields to physical laws

Deadline : 2025-01-12

View details & Apply