MOSIS brings together an interdisciplinary consortium (sociologist, statistician, geographer, data scientist, and expert in multi-agent models) firstly to analyze the evolution of spatial inequalities of intergenerational social mobility in France, secondly to improve statistical and mathematical methods for studying social mobility. From a sociological point of view, the aim is to understand the effects of globalization and technological changes on the spatial variations of social mobility. Thus, this project seeks to study social mobility within the broader context of social, economic, and cultural shifts in French territories by taking into account migratory behaviors. This project has five complementary and interrelated objectives. 1) To combine data from public statistics with digital data, mainly collected on social media. 2) To improve statistical methods for analyzing social mobility. This includes two sub-objectives. The first is to propose a spatial and temporal multilevel specification of the log-linear models traditionally used for analyzing social mobility tables. The second is to demonstrate the advantage of machine learning methods to study social mobility tables, the “object” traditionally studied in quantitative sociology. 3) To identify empirically the relationship between intergenerational social mobility and territorial characteristics and how this relation is affected by globalization and technological changes. 4) To describe the relationship between intergenerational social mobility and spatial mobility, and how this relation impacts the spatial inequalities of social mobility. 5) To build theoretical formal models that explain the link between globalization, technological changes and social mobility and use these models to reproduce observed results and to predict future scenarios. To have a very important amount of individual data so as to be able to analyze social mobility at a fine territorial level, we will combine the most important official data sources on social mobility: the Labor Force Survey, the Permanent Demographic Sample (EDP), the "Déclaration Annuelle de Données Sociales" (DADS) and the survey "Formation and Qualification Professionnelle" (FQP). These data will describe several million individuals. They will be complemented by data on the use of information and communication technologies by companies, and the access to services in different territories (Base Permanente des équipements). Additionally, we will use digital data obtained from Facebook, and LinkedIn advertising to measure detailed socio-demographical, economic and cultural characteristics of territories. Finally, LinkedIn advertising and Chatgpt will be used to collect data on ordinary descriptions of occupations. The interdisciplinary team's first aim is to gain a better understanding of spatial inequalities in social mobility in France by describing the impact of economic, geographical and sociological factors that shape social mobility. The project will also provide the first modern analysis of the relationship between spatial and social mobility in France. This project is inserted in the framework of computational social sciences and aims to renew research on a classic subject in sociology by developing new statistical and mathematical analysis tools. This project will thus enable parallel advances in sociological knowledge and in the fields of mathematics and data science.

Modern statistics and machine learning problems often feature data sitting in an ambient space of high dimension. Yet, methods such as random forests or deep neural networks have recently enabled remarkable performance even in such complex settings. One main reason is that the data can often be explained through a low dimensional structure, hidden to the statistician. In such settings, Bayesian methods such as spike-and-slab variable selection priors, Bayesian additive regression trees (BART), Bayesian deep neural networks or deep Gaussian processes are routinely used by statisticians as well as in physics, astronomy and genomics applications. Among the reasons for the popularity of Bayesian algorithms, one can mention: their flexibility, in that it is relatively easy to model the unknown structure underlying the data through the prior distribution; the broad range of computational methods available, including variational approximations; their ability to quantify uncertainty through so-called credible sets. While there are many empirical successes, there is an important need for understanding and validation for such methods. From the mathematical perspective, one would like to be able to understand and demonstrate under which conditions such algorithms effectively work. The BACKUP project aims at providing theoretical backup for such modern statistical algorithms, around three research avenues. First, new results will be obtained for high-dimensional models and latent variable settings using Bayesian posterior distributions, tackling important recent questions of multiple testing and variable selection. Second, foundational results will be obtained for complex methods such as random forests and Bayesian deep neural networks, both for posteriors and their variational approximations. Third, we will address the fundamental question of uncertainty quantification, by deriving optimal efficient confidence sets from well-chosen Bayesian credible regions.

This project focuses on the effect of stochastic perturbations on oscillatory phenomena in dynamical systems. Oscillations are present in a vast number of systems in physics, biology and chemistry. Noise acting on these systems may drastically modify the oscillation patterns, or, in the case of excitable systems, create oscillations that were absent in the unperturbed system. Systems displaying oscillations are by essence irreversible, and therefore the mathematical theory of their stochastic perturbations is still in its infancy. Recently a number of new mathematical techniques have emerged that promise substantial progress in the description of non-reversible systems. The aim of this project is to develop these methods further and to combine them in order to obtain effective tools for the study of oscillations in stochastic systems. These tools will be applied to the description of oscillations in dusty plasmas, and to several models originating in mathematical biology.

The project is devoted to the development of statistical methods specifically designed for analysing different types of ecological networks: trophic, mutualistic, competition or antagonistic and host-parasite systems. We propose to create a unique consortium of researchers combining applied statisticians with long-standing experience on life-science modeling and ecologists at the forefront of their domain to tackle the challenges posed by the advanced statistical modeling of ecological networks. Our proposal includes 1) integrating space and time dimensions in ecological networks modeling and developing tools for the comparison of networks along environmental gradients; 2) integrating multiple types of interactions, taking advantage of covariate information (such as species traits, distributions, phylogenies and environmental variables) when available; 3) incorporating sampling effects in our analyses and 4) providing predictions on ecosystem response to environmental changes.

The international market of Rungis is one of the most sophisticated market places in the world from an operational perspective. Over a million different food products circulate through this incredible platform: 2.82 million tonnes of fresh produce pass through the market each year, close to of 9 billion euros in turnover. However, the market lags behind in technological terms and is struggling to digitalize itself. This is the observation made by Califrais, (foodtech startup founded in 2014), which offers an innovative technological solution to optimize the flow of goods (in this case of fresh produce) with end-to-end control of the value chain and tailor-made IT tools. (Sourcing, product expertise, logistics and inventory management.) The service was initially aimed at HoReCa professionals, and in 2020 was extended to include all retail customers (Foodufrais) offering them an access to the high quality fresh produce straight from the market of Rungis. In order to guarantee an exceptional level of service while supporting strong growth, Califrais uses the latest breakthroughs in artificial intelligence (AI) and big data.3 Whether it is logistics or customer relationship management, recent techniques resulting from these cutting-edge technologies tailored to the peculiarities of the functioning of the market of Rungis will indeed gradually revolutionize the level of service, as we know it today. On one hand, by offering personalized recommendation and management strategies for customers on the platform, and also by optimizing all the internal operational processes of wholesellers on the other hand. The objective being to reduce spoilage and food waste, while reducing the overall purchasing costs for the customer. The modeling of complex flows of large-scale foodstuffs (changes in product prices, demand, etc.) will make it possible to optimize the wholesalers inventory, and thus help them anticipate demand and limit waste. It will optimize last mile logistics, considerably reducing Parisian traffic from Rungis; and also bring about predictive decision support systems and intelligent customer service. The Laboratory of Probability, Statistics and Modeling (LPSM) and Califrais have decided to join forces within the framework of LabCom LOPF (Large-scale Optimization of Product Flows) to answer these questions, which fit perfectly into the research themes of the “Statistics, data, algorithms” team. The LPSM is in fact particularly suitable since it covers, all cutting-edge themes involving statistics, while opening up to the cross-cutting challenges posed by digital technology, such as intensive computing or machine learning. The scientific issue of LabCom is to bring the technological developments initiated by Califrais to the scale of market of Rungis by 2023. The main objective is to develop innovation and applied research activity in the field of: optimization of the flow of large-scale fresh products, taking into account all the operational constraints specific to the market (supply chain, sell-by-date of products, multi-site storage, etc.). The modeling challenges are technically ambitious but realistic. The dynamism of the Califrais in the field of innovation, as well as its unique expertise in the subject, coupled with the recognized excellence in theoretical and applied research of the LPSM will make it possible to provide concrete and rapidly operational solutions.