SENELLART Pierre

Data management

pierre.senellart [at] ens.fr / Twitter: @PierreSenellart

Pierre Senellart dav

Short bio

Professor and deputy director, Department of Computer Sicence, École normale supérieure. Head of the Valda Inria team. Adjunct professor, Télécom Paris. Research fellow at the Centre on regulation in Europe.

Topics of interest

Data management, uncertainty, provenance, large-scale networks

Project in Prairie

Pierre Senellart’s research deals with management and mining of large-scale structured and semi-structured data in general, and largescale networks in particular, addressing issues such as management of the provenance, privacy, and uncertainty of data, and scalability through leveraging the structure of data. He is a co-coordinator of the PSL graduate program on data science and AI, in particular teaching within the PSL IASD Master.

Quote

Uncertain data are pervasive in artificial intelligence systems: input data are often imprecise, incomplete, or even contradictory, while automatic tools run on these data often produce imperfect annotations or predictions. A major challenge is to properly manage this uncertainty in data, keeping track of the confidence in individual data items throughout complex processes. For explicability and traceability purposes, it is also important to keep track of the provenance of data: where it comes from, how it was produced, etc. Uncertainty and provenance management are two of the main research issues in modern data management.

MASSOULIÉ Laurent

Networks and Machine Learning

laurent.massoulie [at] inria.fr

Laurent Massoulie

Short bio

Research Director at Inria, Director of the Microsoft Research-Inria Joint Centre, and Professor at the Applied Mathematics Centre of Ecole Polytechnique. Co-author of the Best Paper Award-winning papers of IEEE INFOCOM 1999, ACM SIGMETRICS 2005, ACM CoNEXT 2007, NeurIPS 2018. Elected a Technicolor Fellow in 2011. Recipient of the “Grand Prix Scientifique” from the Del Duca Foundation delivered by the French Academy of Sciences in 2017.

Topics of interest

Unsupervised learning, distributed machine learning, modeling and algorithmic design for distributed systems and networks

Project in Prairie

Laurent Massoulié will develop distributed algorithms for learning from data spread over several machines, to efficiently exploit communication resources between data locations, and storage and compute resources at data locations. He will develop efficient algorithms for unsupervised learning from ‘graphical data’. He will also address fairness and privacy challenges of machine learning, in particular in the contexts of recommender systems and matching markets.

Quote

Relational, or ‘graphical’ data is becoming ubiquitous (e.g. social / biological / transportation networks, energy grids…). Its treatment calls for new methods to construct and process adequate representations of data points in suitable spaces. There are many important scenarios where data must be distributed on several network locations, e.g. when it is too large to fit on a single machine, or when it can’t leave administrative boundaries due to privacy concerns. New distributed algorithms, and possibly new network architectures are needed for efficient learning from data distributed over a network.

LELARGE Marc

Representation learning

marc.lelarge [at] inria.fr

Marc Lelarge

Short bio

Dr. Marc Lelarge is a researcher at INRIA. He is also a lecturer in deep learning at Ecole Polytechnique (Palaiseau, France) and Ecole Normale Superieure. He graduated from Ecole Polytechnique, qualified as an engineer at Ecole Nationale Superieure des Telecommunications (Paris) and received a PhD in Applied Mathematics from Ecole Polytechnique in 2005. Dr. Marc Lelarge received the NetGCoop 2011 Best Paper Award with his PhD student E. Coupechoux, was awarded the 2012 SIGMETRICS rising star researcher award and the 2015 Best Publication in Applied Probability Award with Mohsen Bayati and Andrea Montanari for their work on compressed sensing.

Topics of interest

Machine learning, deep learning, graphs and data analytics

Project in Prairie

Marc Lelarge’s research focuses on representation learning with an emphasis on explainability. Marc Lelarge develops algorithms for learning and mining from relational and graphical data and pursue collaborations in medical informatics (active learning and NLP for electronic health records). Marc Lelarge is teaching courses on deep learning at École Normale Supérieure and École Polytechnique.

Quote

Many scientific fields study data with an underlying graph or manifold structure such as social networks, sensor networks, biomedical knowledge graphs. The need for new optimization methods and neural network architectures that can accommodate these relational and non-Euclidean structures is becoming increasingly clear. Indeed, the statistical properties of data defined on high-dimensional graphs are very different form the stationarity, locality and compositionality assumptions at the core of the success of deep learning models in vision or NLP. For such type of data, I am working on a recent deep learning approach: graph neural networks able to learn a message passing algorithm and aggregation procedure to compute an embedding of the graph and its nodes.

BHARGAVAN Karthikeyan

Security

karthikeyan.bhargavan [at] inria.fr

Karthikeyan Barghavan

Short bio

Research director, Inria Paris (2015-). Researcher, Inria Paris (2009-2015). Researcher, Microsoft Research UK (2004-2009). Recipient of ERC Consolidator Grant (2016) and ERC Starting Grant (2010). Awarded EC Horizon Impact Award (2019), Prix Jeune Chercheur Inria–Académie des sciences (2016), Microsoft Research Outstanding Collaborator Award (2016), Levchin Prize for Real World Cryptography (2016).

Topics of interest

Programming Languages, Software Verification, Applied Cryptography, Security Protocol Design and Analysis

Project in Prairie

Karthikeyan Bhargavan’s research lies at the intersection of programming languages, software verification, and applied cryptography. He will work on developing efficient cryptographic protocols and high-assurance software for privacy-preserving machine learning algorithms. In collaboration with AI researchers, he seeks to design new programming language abstractions and verification tools that can enable developers to build formally verified AI applications that provide built-in security and privacy guarantees. He will also teach courses on cryptographic protocols and high-assurance software development at MPRI and PSL.

Quote

Machine learning systems are notoriously hungry for data, and often this
data is personal and private. How do we then build impactful AI applications that can be used in sensitive areas like medicine and transportation without violating the privacy of our users? The answer lies in building in security and privacy enhancing mechanisms as essential components of AI systems and formally proving that these mechanisms meet strong security requirements. In my team, we design programming languages and verification tools that can help build high-assurance security-oriented software. Our next research challenge is to create provably secure software that also learns, but preserves user privacy.