Visitez notre page








Les codes et logiciels de l'unité sont hébergés à divers endroits selon leur niveau d'avancement, les besoins de confidentialité et le besoin de diffusion.

On trouve notamment ici Ci-dessous, nous pointons uniquement quelques uns des codes et logiciels les plus avancés de l'unité, répertoriés par ses membres.

Quelques codes de l'UMR

jinns : physics-informed neural network with JAX

The jinns python package allows to solve differential equations -ODEs,  PDEs, or systems of thereof - using the framework of physics-informed neural network and the JAX ecosystem (automatic differentiation, just-in-time compilation, and more). Its interface is focused on user flexibility, allowing to easily define a "physics" loss and handling the rest. The development focuses on inverse problems, where one wishes to infer some of the differential equation parameters using observed data   Many examples notebooks are available in the package documentation :


Developers & Maintainers : Hugo Gangloff, Nicolas Jouvin

Multivariate Poisson lognormal models

The Poisson lognormal model and variants can be used for a variety of multivariate problems when count data are at play (including PCA, LDA and network inference for count data). The R package `PLNmodels` and the Python port `pyPLNmodels` implements efficient algorithms to fit such models accompanied with a set of functions for visualization and diagnostic.


shinySbm is a R package containing a shiny application. This application provides a user-friendly interface for network analysis based on the sbm package made by Chiquet J, Donnet S and Barbillon P (2023) CRAN. The sbm package regroups into a unique framework tools for estimating and manipulating variants of the stochastic block model. shinySbm allows you to easily apply and explore the outputs of a Stochastic Block Model without programming. It is useful if you want to analyze your network data (adjacency matrix or list of edges) without knowing the R language or to learn the basics of the sbm package.

Stochastic block models (SBMs) are probabilistic models in statistical analysis of graphs or networks, that can be used to discover or understand the (hidden/latent) structure of a network, as well as for clustering purposes.

Stochastic Block Models are applied on network to simplify the information they gather, and help visualize the main behaviours/categories/relationships present in your network. It’s a latent model which identify significant blocks (groups) of nodes with similar connectivity patterns. This could help you to know if your network: hides closed sub-communities, is hierarchical, or has another specific structure.

With shinySbm you should also be able to:

  • Easily run a Stochastic Block Model (set your model, infer associated parameters and choose the number of blocks)
  • Get some nice outputs as matrix and network plots organized by blocks
  • Get a summary of the modelling
  • Extract lists of nodes associated with their blocks

Links :