Confiance.ai: Trustworthy AI for Critical Systems >>Learn more
Confiance.ai: Trustworthy AI for Critical Systems >>Learn more
Statistical Learning
Generative (Latent Variable) Models
Unsupervised learning
High-dimensional and Large-Scale scenarios
Applications
Associate Editor of Neurocomputing - Elsevier
I have the privilege to be involved since Sept.2022 in Confiance.AI, the French AI flagship program to industrialize trustworthy AI-based critical systems, coordinated by IRT SystemX with Bertrand Braunschweig the scientific coordinator. In AI systems, including machine learning based AI, it is crucial to guarantee key properties like accuracy, robustness, explainability, fairness, privacy, among many other primary values of the Trust. Trust is the key objective for deploying AI-based systems in accordance with the European values, as defined by the European commission and in preparation of the forthcoming AI Act. In industry, Confiance.AI, the largest technological research program in the French AI strategy, as an industry-oriented program, has the strong ambition of breaking down the scientific and technological barriers associated with the development of trustworthy AI-based systems of different criticality levels, and their deployment and industrialisation in partnership with pilot industrial actors. To this end, the program has an original and integrated research and development validation strategy based on industrial use cases, covering a broad spectrum of critical systems, and endowed with major research advances in the AI landscape, including - but not limited to - data-based AI solutions and deep neural networks, with application on industrial use cases covering image processing, time series and structured data, video, audio and text data, as well as the introduction of other AI formalisms including knowledge-based and hybrid approaches. >>Learn more
Large-scale data analysis is an inherently multidisciplinary area and is becoming of broader interest for today's society. The ANR project SMILES is a collaborative fundamental research project that aims at introducing an unsupervised statistical modeling framework and scaled inference algorithms for transforming large-scale data into knowledge. It considers the large-scale context as a whole, with its main issues related to inference from a big volume of data of very high dimension and underlying complex hidden structures. The key tenet of SMILES is to introduce large-scale latent data models for unsupervised data classification and large-scale regression-based sparse (non)parametric models for data representation. The knowledge extraction will namely consist in automatically retrieving hidden structures, summarizing prototypes, groups, sparse representations. We consider different data settings, including functional data, multimodal bioacoustical data, and biological data. >>Learn more
AStERiCs Est un projet de recherche fondamentale financé dans le cadre du dispositif RIN (Réseaux d'Intérêts Normands)-Recherche dont l'objectif structurel est de fédérer la recherche scientifique en Normandie dans le domaine de la science statistique des données, en s'appuyant sur une démarche scientifique pluridisciplinaire impliquant modélisation mathématique, inférence, représentation et classification de données issues d'environnements complexes, hétérogènes, dynamiques et incertains. AStERiCs vise à élaborer un cadre scientifique et technique, complet, pour traiter, analyser, exploiter et valoriser des données massives, complexes, hétérogènes, dynamiques et peu ou non-annotées. Le but est de transformer des données en connaissances sous forme de représentations précises des informations liées aux données, de catégorisations pertinentes de telles informations, jusqu’à la valorisation de celles-ci en révélant/restaurant le modèle générateur des données. Le projet AStERiCs traite ainsi le problème de la grande échelle, sous tous ses aspects de modélisation et d'inférence. Plus précisément, les axes de recherche traitent des grands thèmes suivants : Statistique, Apprentissage, Analyse de données, Classification, Optimisation, Traitement du signal, Grande dimension. >>Learn more
The Research Summer School on Statistics for Data Science (S4D) is a multidisciplinary international summer school at the interface of several disciplines: statistics, computer science (machine learning), optimization and signal processing, with a focus on probabilistic/statistical formulations of recent methodologies dedicated to the problem of learning from large-scale complex data. It involves, during a week, for an audience of PhD students, engineers, postdocs and scientists, several international pioneering scientists in modern statistics, machine learning and artificial intelligence, to present their expertise on both theoretical and technical aspects of models and algorithms to address data science problems with a focus on large-scale/high-dimensional scenarios. The objectives of the talks are to: - present the basic developments of models for phd students and practitioners to understand the basics of the domain; - highlight for students, scientists and practitioners the major scientific issues necessary for the development of such models and on the related theoretical and computational problems; - present and analyze the latest theoretical advances in modeling and practical considerations regarding statistical learning, classification, representation, of large-scale raw data, in various application area >>Learn more
