{"id":663087,"date":"2021-01-25T05:52:45","date_gmt":"2021-01-25T13:52:45","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-group&p=663087"},"modified":"2024-03-06T09:35:10","modified_gmt":"2024-03-06T17:35:10","slug":"inria-joint-centre","status":"publish","type":"msr-group","link":"https:\/\/www.microsoft.com\/en-us\/research\/collaboration\/inria-joint-centre\/","title":{"rendered":"Inria Joint Center"},"content":{"rendered":"

\"INRIAThe Microsoft Research-Inria Joint Center, Inria JC, is a collaborative research partnership between Microsoft Research and Inria<\/a>, the French Public Research Institute in Computer Science. Since its creation in 2005, the Inria JC has been home to\u00a0over\u00a025\u00a0projects co-led by researchers from\u00a0Inria\u00a0and MSR. There are currently eleven active projects, addressing a wide range of cutting-edge research challenges.<\/p>\n

Security is an area of particular focus, with projects on cryptography for blockchains,\u00a0certification of distributed algorithms\u2019 correctness, and\u00a0design of cryptographic protocols for the Internet with certified security properties.<\/p>\n

Artificial Intelligence is another main focus of the Inria JC with projects on machine learning and computer vision (specifically for Mixed Reality devices such as HoloLens) and alternative paradigms to reinforcement learning for continuously evolving and interacting agents.\u00a0\u00a0In Computer Vision, the Inria JC partners closely with the Mixed Reality & AI Zurich Lab<\/a>.<\/p>\n

\"Photo\u201cCollaboration between computer science researchers from academia and industry has never made more sense than today, given the exciting challenges and opportunities offered by our field. The MSR-Inria JC empowers such collaborations between fellow researchers at\u00a0Inria\u00a0and Microsoft Research on topics from artificial intelligence to security and quantum computing.\u201d\u00a0 <\/em><\/p>\n

Laurent Massoulie, Inria JC Managing Director<\/p>\n

The Inria JC is governed by a Management Committee consisting of representatives of Inria, Microsoft Research and Microsoft France. Among other things, the Management Committee oversees the progress of existing projects and identifies new project opportunities.<\/p>\n

Address<\/h3>\n

Microsoft Research – Inria JC
\n2 rue Simone Iff
\n75012 PARIS<\/p>\n","protected":false},"excerpt":{"rendered":"

The Centre’s objective is to pursue fundamental, long-term research in formal methods, software security, and the application of computer science research to the sciences.<\/p>\n","protected":false},"featured_media":699334,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":true,"_classifai_error":"","msr_group_start":"","footnotes":""},"research-area":[13556,13562,13546,13553,13558,13547],"msr-group-type":[243721],"msr-locale":[268875],"msr-impact-theme":[],"class_list":["post-663087","msr-group","type-msr-group","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-research-area-computer-vision","msr-research-area-computational-sciences-mathematics","msr-research-area-medical-health-genomics","msr-research-area-security-privacy-cryptography","msr-research-area-systems-and-networking","msr-group-type-collaboration","msr-locale-en_us"],"msr_group_start":"","msr_detailed_description":"","msr_further_details":"","msr_hero_images":[],"msr_research_lab":[],"related-researchers":[{"type":"guest","display_name":"Matthieu Armando","user_id":665112,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Balthazar Bauer","user_id":665160,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Ioannis Filippidis","user_id":717796,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Florian Groult","user_id":717787,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Yana Hasson","user_id":608811,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Hadrien Hendrikx","user_id":665097,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Antoine Plouviez","user_id":665157,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Marina Polubelova","user_id":717790,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"R\u00e9my Portelas","user_id":717811,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Mathieu Rita","user_id":717820,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Maria Laetitia Teodorescu","user_id":717814,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Boyao Zhou","user_id":665118,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Dimitri Zhukov","user_id":665133,"people_section":"Students and Postdocs","alias":""},{"type":"guest","display_name":"Antoine Defourn\u00e9\u00a0\u00a0","user_id":717799,"people_section":"Students and Postdocs","alias":""},{"type":"user_nicename","display_name":"Melissa Chase","user_id":32878,"people_section":"Microsoft - Principal Investigators","alias":"melissac"},{"type":"user_nicename","display_name":"Antoine Delignat-Lavaud","user_id":31056,"people_section":"Microsoft - Principal Investigators","alias":"antdl"},{"type":"user_nicename","display_name":"C\u00e9dric Fournet","user_id":31819,"people_section":"Microsoft - Principal Investigators","alias":"fournet"},{"type":"user_nicename","display_name":"Esha Ghosh","user_id":37851,"people_section":"Microsoft - Principal Investigators","alias":"esghosh"},{"type":"user_nicename","display_name":"Chris Hawblitzel","user_id":31425,"people_section":"Microsoft - Principal Investigators","alias":"chrishaw"},{"type":"user_nicename","display_name":"Katja Hofmann","user_id":32468,"people_section":"Microsoft - Principal Investigators","alias":"kahofman"},{"type":"user_nicename","display_name":"Markus Kuppe","user_id":32784,"people_section":"Microsoft - Principal Investigators","alias":"makuppe"},{"type":"user_nicename","display_name":"Leslie Lamport","user_id":32614,"people_section":"Microsoft - Principal Investigators","alias":"lamport"},{"type":"user_nicename","display_name":"Jay Lorch","user_id":32732,"people_section":"Microsoft - Principal Investigators","alias":"lorch"},{"type":"user_nicename","display_name":"Marc Pollefeys","user_id":36191,"people_section":"Microsoft - Principal Investigators","alias":"mapoll"},{"type":"user_nicename","display_name":"Jonathan Protzenko","user_id":33286,"people_section":"Microsoft - Principal Investigators","alias":"protz"},{"type":"user_nicename","display_name":"Tahina Ramananandro","user_id":36293,"people_section":"Microsoft - Principal Investigators","alias":"taramana"},{"type":"user_nicename","display_name":"Aseem Rastogi","user_id":36021,"people_section":"Microsoft - Principal Investigators","alias":"aseemr"},{"type":"user_nicename","display_name":"Srinath Setty","user_id":33709,"people_section":"Microsoft - Principal Investigators","alias":"srinath"},{"type":"user_nicename","display_name":"Sudipta Sinha","user_id":33748,"people_section":"Microsoft - Principal Investigators","alias":"sudipsin"},{"type":"user_nicename","display_name":"Paul Smolensky","user_id":36353,"people_section":"Microsoft - Principal Investigators","alias":"psmo"},{"type":"user_nicename","display_name":"Nikhil Swamy","user_id":33138,"people_section":"Microsoft - Principal Investigators","alias":"nswamy"},{"type":"user_nicename","display_name":"Santiago Zanella-B\u00e9guelin","user_id":33518,"people_section":"Microsoft - Principal Investigators","alias":"santiago"},{"type":"guest","display_name":"Martin de la Gorce","user_id":735115,"people_section":"Microsoft - Principal Investigators","alias":""},{"type":"guest","display_name":"Francis Bach","user_id":665040,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Karthikeyan Bhargavan","user_id":571929,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Edmond Boyer","user_id":665049,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Damien Doligez","user_id":665163,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Emmanuel Dupoux","user_id":717817,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Jean-S\u00e9bastien Franco","user_id":665052,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Ivan Laptev","user_id":665055,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Laurent Massoulie","user_id":665043,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Stephan Merz","user_id":665169,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Pierre-Yves Oudeyer","user_id":717808,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"David Pointcheval","user_id":664998,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Josef Sivic","user_id":665061,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Stefano Zacchiroli","user_id":717805,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Roberto Di Cosmo","user_id":717802,"people_section":"Inria - Principal Investigators","alias":""},{"type":"guest","display_name":"Fran\u00e7ois Cuny","user_id":717412,"people_section":"Management Board","alias":""},{"type":"guest","display_name":"\u00c9ric Fleury","user_id":718546,"people_section":"Management Board","alias":""},{"type":"guest","display_name":"Jean-Fr\u00e9d\u00e9ric Gerbeau","user_id":717415,"people_section":"Management Board","alias":""},{"type":"guest","display_name":"Eneric Lopez","user_id":743158,"people_section":"Management Board","alias":""},{"type":"guest","display_name":"Laurent Massoulie","user_id":665043,"people_section":"Management Board","alias":""},{"type":"guest","display_name":"Bernard Ourghanlian","user_id":717403,"people_section":"Management Board","alias":""},{"type":"user_nicename","display_name":"Evelyne Viegas","user_id":31768,"people_section":"Management Board","alias":"evelynev"}],"related-publications":[339641,720403,866301,392960,720412,619023,720418,640176,720424,655743,720430,662424,720436,671178,720442,701437,720448,152126,720364,725464,161413,720376,725473,328493,720391,866286],"related-downloads":[736291],"related-videos":[],"related-projects":[],"related-events":[],"related-opportunities":[],"related-posts":[],"tab-content":[{"id":0,"name":"Projects","content":"

Current projects<\/h2>\r\n[accordion]\r\n\r\n[panel header=\"Acquisition and Synthesis of High-quality Four-dimensional Data\"]\r\n\r\nInria PI:<\/strong> Edmond Boyer, Jean-S\u00e9bastien Franco\r\n\r\nMS PI:<\/strong> Federica Bogo, Martin de la Gorce\r\n\r\nPhD students:<\/strong> Matthieu Armando, Boyao Zhou\r\n\r\nOne aim of this project is the synthesis of 3D-shape models based on observations from several viewpoints. A challenge to reach this aim is to remove noise from acquired shape models. To that end we develop denoising algorithms based on graph convolutional neural networks.\r\nA second aim of this project is the completion of shape and motion information from partial observations, e.g. sparse point clouds. To address it we develop approaches based on Gaussian processes combined with deep learning.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Autonomous Learning \"]\r\n\r\nInria PI:<\/strong> Pierre-Yves Oudeyer\r\n\r\nMSR PI:<\/strong> Katja Hofmann\r\n\r\nPhD students:<\/strong> R\u00e9my Portelas, Laetitia Teodorescu\r\n\r\nThis project takes place in the context of developing the foundations of a novel machine learning framework, called Intrinsically Motivated Goal Exploration (IMGEPs), focused on the challenge of autonomous learning of diverse repertoires of skills in open and changing environments. This is an alternative to the classical reinforcement learning framework, which is currently reaching both conceptual and practical limits. It is grounded in a decade of work modelling mechanisms of curiosity-driven learning and development in human infants. In particular, it aims to develop algorithms enabling i) to learn incrementally goal spaces representations and associated internal goal achievement reward function, with the aim to foster efficient discovery of a diversity of skills; ii) a social peer to guide curiosity-driven learners using natural language, both fostering more efficient exploration leveraging language input, and enabling external users to leverage skills learned autonomously by the machine by using language-based queries.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Cryptography for the Blockchain \"]\r\n\r\nInria PIs:<\/strong> David Pointcheval\r\n\r\nMSR PIs:<\/strong> Melissa Chase, Esha Ghosh, Santiago Zanella\r\n\r\nPhD students:<\/strong> Balthazar Bauer, Antoine Plouviez\r\n\r\nThe project aims to advance state-of-the-art in blockchain technology. Blockchains as exemplified by Bitcoin, Ethereum, ZCash, show promise to enable guaranteed transactions without the need for a central authority. Among identified objectives, researchers of the project are looking at: making blockchains \u201cgreener\u201d, for instance by replacing notion of proof of work that underlies bitcoin by \u201cproof of storage\u201d, and develop new applications of blockchains, in the spirit of recent proposals to perform decentralized verification of e.g., web certificates, and public-key infrastructure system.\r\n\r\n[\/panel]\r\n[panel header=\"Flexible AI\"]\r\n\r\nInria PI:<\/strong> Emmanuel Dupoux\r\n\r\nMSR PI:<\/strong> Paul Smolensky\r\n\r\nPhD student:<\/strong> Mathieu Rita\r\n\r\nRecent progresses in deep learning led to the successful development of neural networks able to achieve complex tasks in domains such as computer vision or natural language. These successes are attained by networks that are trained with massive amounts of data. However, the interactive and functional aspects of intelligence are almost completely ignored. Indeed, researchers only begin to take an interest in introducing communicating tools in multi-agent settings. This research aims in studying the conditions under which populations of agents develop a communication that helps them to succeed in their tasks. From a scientific perspective, understanding the evolution of languages in communities of deep agents and its emergent properties can shed light on human language evolution. From an applied perspective, endowing deep networks with the ability to solve problems interactively by communicating with each other and with us should make them more flexible and useful in everyday life.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Machine Learning for Distributed Environments \"]\r\n\r\nInria PIs:<\/strong> Francis Bach, Laurent Massouli\u00e9\r\n\r\nMSR PIs:<\/strong> S\u00e9bastien Bubeck\r\n\r\nPhD Student:<\/strong> Hadrien Hendrikx\r\n\r\nThe project aims to develop efficient distributed ML algorithms suited to run for instance in the cloud.\r\n\r\nLarge-scale machine learning algorithms, whether supervised or unsupervised, are ideally run on a cloud platform with many cores organized in a hierarchical fashion in datacenters. Important efficiency gains can be had by designing new algorithms tailored to exploit such distributed platforms. Specific objectives include design of new algorithms for supervised learning (e.g., distributed, accelerated gradient descent) and unsupervised learning (e.g., distributed spectral analysis and graph clustering)\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Project Everest\"]\r\n\r\nInria PI:<\/strong> Karthik Bhargavan\r\n\r\nMSR PIs:<\/strong> Antoine Delignat-Lavaud, C\u00e9dric Fournet, Jonathan Protzenko, Nikhil Swamy, Santiago Zanella\r\n\r\nPostdoc:<\/strong> Florian Groult\r\n\r\nPhD Student:<\/strong> Marina Polubelova\r\n\r\nThe project aims at building formally verified high-performance standard-compliant cryptographic components for securing Internet communications. These components can be used either as drop-in replacements for existing libraries and protocols, or to build verified secure sub-systems and applications. By construction, these components prevent flaws and vulnerabilities that litter existing implementations and require frequent security patches\u2014see, e.g., the 3SHAKE, FREAK and LOGJAM attacks uncovered as part of our research. More details can be found at the following sites:\r\n\r\nView project page<\/a>\r\n\r\nhttps:\/\/project-everest.github.io\/<\/a>\r\n\r\nhttps:\/\/www.fstar-lang.org\/<\/a>\r\n\r\nhttps:\/\/mitls.org\/<\/a>\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Quantum Error Correction\"]\r\n\r\nInria PI:<\/strong> Anthony Leverrier\r\n\r\nMS PIs:<\/strong> Nicolas Delfosse\r\n\r\nQuantum hardware suffers from much higher error rates than classical devices. As a result, extensive error correction is necessary to execute a large scale quantum algorithm. Quantum LDPC codes could lead to a significant reduction the cost of quantum error correction but they need better decoders. With this project, we will explore the potential of the Union-Find decoder to decode quantum LDPC codes.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Quantum Walks and Algorithms\"]\r\n\r\nInria PI:<\/strong> Alain Sarlette\r\n\r\nMS PIs:<\/strong> Martin Roetteler, Matthias Troyer\r\n\r\nPostdoc:<\/strong> Matthys Rennela\r\n\r\nThe aim of this project is to accurately evaluate the resources required for implementing quantum speedup based on the \"quantum fast forwarding\" (QFF) algorithm for Markov chains. The QFF routine indeed operates at an intermediate level between a graph-based quantum oracle, and an output to be extracted from the generated Markov chain distribution. We will consider concrete applications to specify these oracle and output layers, evaluate the complexity and opportunities for implementing them in a quantum computer, and thereby propose a full-chain quantum computation with graph oracles.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Software Heritage and Software Productivity \"]\r\n\r\nInria PIs:<\/strong> Roberto di Cosmo, Stefano Zacchiroli\r\n\r\nMSR PI:<\/strong> Tom Zimmermann\r\n\r\nSoftware Heritage is today the largest corpus of software source code artifacts, containing over 6 billion unique source code files, 1 billion commits, coming from more than 90 million software projects. So far, empirical studies of software engineering have been conducted on much smaller repositories, and as such can\u2019t be fully conclusive. The aim of this project is to reproduce previous empirical studies relevant for the improvement of developer productivity on Software Heritage. In particular, regarding Software Heritage it will analyze: the extent of large-scale code reuse in open-source software; identification of topics in source code via semantic clustering; relation between programming language and code quality; quality and productivity outcomes relating to continuous integration.\r\n\r\n[\/panel]\r\n\r\n[panel header=\"TLA Proof System \"]\r\n\r\nInria PIs: <\/strong>Damien Doligez, Stephan Merz\r\n\r\nMSR PIs:<\/strong> Markus Kuppe, Leslie Lamport\r\n\r\nPostdoc:<\/strong> Ioannis Filippidis\r\n\r\nPhD student:<\/strong> Antoine Defourn\u00e9\r\n\r\nThe project aims to develop a proof assistant for enabling certification of TLA+-expressed specification.\r\n\r\nTLA+, with the associated Pluscal algorithm language, have been proposed by Leslie Lamport as tools for developing and verifying algorithms for concurrent and distributed systems. These have gradually matured and have recently been used for industrial applications. They are now entering mainstream use for certifying properties of distributed systems. An associated proof assistant is an important complement to model-checking for establishing properties of TLA+-expressed algorithm specifications. More details can be found at: https:\/\/lamport.azurewebsites.net\/tla\/tla.html<\/a>\r\n\r\nhttps:\/\/tla.msr-inria.inria.fr<\/a>\r\n\r\n[\/panel]\r\n\r\n[panel header=\"Video Understanding for Personal Assistants \"]\r\n\r\nInria PIs: <\/strong>Ivan Laptev, Josef Sivic\r\n\r\nMS PIs:<\/strong> Marc Pollefeys, Johannes Sch\u00f6nberger\r\n\r\nPhD student:<\/strong> Yana Hasson, Dimitri Zhukov\r\n\r\nThe project aims to develop tools for the HoloLens Mixed Reality device, in particular solutions for construction of visual cues to be superimposed to real scene in order to help conduct practical tasks (such as changing a car tyre). Ongoing work at Inria shows promise on automatic processing of YouTube videos of instructions for various tasks to identify key stages and visual signatures of these steps. It is a bold challenge to try and go from there to the actual production of visual cues to be projected onto HoloLens to assist users in conducting chosen tasks.\r\n\r\n[\/panel]\r\n\r\n[\/accordion]\r\n

<\/h2>\r\n

Archived projects<\/h2>\r\n[accordion]\r\n\r\n[panel header=\"4D Cardiac MR Images\"]\r\n\r\n\"4D\r\n\r\nThis project finished in 2013 \u2013 then continued by \u201cMedilearn\u201d<\/em>\r\n

Goals of the project<\/h3>\r\nGiven a large database of cardiac images of patients stored with an expert diagnosis, we wish to automatically select the most similar cases to the cardiac images of a new patient.\r\n

Application<\/h3>\r\nThis is important, for instance to try and estimate the optimal time for an operation when a Tetralogy of Fallot condition is diagnosed (see fig. 1).\r\n

Science<\/h3>\r\nWe want to be able to index images based on e.g. the shape of the heart, the dynamics of the myocardium, or the presence of anomalies. This would require capturing the right level of shape, motion and appearance characteristics in a compact and efficient way so as to enable fast indexation and retrieval even for hundreds of 4D datasets. Using state of the art efficient machine learning techniques is also of paramount importance. The design of both the visual features and the classification algorithms have to be informed by medical experts so as to make sure the final system remains relevant from a clinical point of view.\r\n\r\nSuch a system would provide a strong innovation in cardiology, a learning tool for residents and young physicians, and a new tool to help physicians to assess a diagnosis and plan a therapy. The method could emphasize local motion and\/or local shape singularities (e.g. septal flash (motion anomaly) vs. ventricle overload (shape anomaly), with the possibility to actually weight motion vs. shape features.\r\n

The MSRC\u00ad-Inria collaboration<\/h3>\r\nKey points to succeed in developing such a system are the following:\r\n