Trois journées par an (Lyon, Marseille et Grenoble) avec 4 exposés.
Journée 3 à Grenoble le 21 juin 2024 (10h-16h)
Inscription en suivant le lien suivant: https://lite.framacalc.org/4vaanuereg-a7j2.
La journée aura lieu dans la salle 4 à l’Institut Fourier.
- 10h00 – 10h30: Accueil avec viennoiseries
- 10h30 – 11h30: Julien Chevallier (LJK, Grenoble), Estimation de la densité d’un graphe de dépendance en grande dimension
Abstract :
- 11h30 – 12h30: Patrick Vallet (INRAE, Grenoble)
Abstract : TBA
- 12h30 – 13h45 : Déjeuner (buffet)
- 13h45 – 14h45: Clara Lage (ENS de Lyon): Identifying a piecewise affine signal from its nonlinear observation – application to DNA replication analysis
Abstract : An important challenge in DNA replication analysis is to recover a so-called timing profile, that contains important information about the replication dynamics, from nonlinear observations. We show that this can be expressed as a nonlinear inverse problem where the unknown timing profile can be assumed to be piecewise affine. As this problem cannot be directly addressed with techniques for linear inverse problems, we propose a novel formalism and computational approach to harness it. In the noiseless case, we establish sufficient identifiability conditions for the timing profile, and prove that it is the solution of a non-convex optimization problem. These problems are specially challenging because of their multiple local minima. We propose the DNA-Inverse optimization method that is capable of finding their global solution in the noiseless case and proved to be effective in numerical experiments for noisy signals. Comparative analysis against state-of-the-art optimization methods highlights the superior computational efficiency of our approach. The method enables the automatic recovery of all configurations of the replication dynamics, crucial for DNA replication analysis, which was not possible with previous methods.
- 14h45 – 15h: Pause avec gâteau
- 15h – 16h: Vincent Miele (CNRS, LECA Chambéry/Grenoble): ELGRIN, un modèle statistique pour quantifier l’effet des interactions biotiques sur la distribution des espèces le long de gradients environnementaux.
Abstract : We propose a novel statistical model, called ELGRIN (in reference to Charles Elton and Joseph Grinnell) that can handle the effects of both environmental factors and known interspecific interactions (aka a metanetwork) on species distributions. We rely on Markov random fields to handle dependencies between variables using a graph. More specifically, ELGRIN jointly models the presence and absence of all species in a given area in function of environmental covariates and the topological structure of the known metanetwork. It separates the interspecific interaction effects from those of the environment on species distributions.
Using various simulated and empirical data, we demonstrate the suitability of ELGRIN to address the objectives for various types of interspecific interactions like mutualism, competition and trophic interactions. We then apply the model on vertebrate trophic networks in the European Alps.
