Séminaire Statistique
organisé par l'équipe Statistique
-
Antoine Heranval
Analyzing temporal dependence between extreme events using point processes
10 avril 2026 - 11:00Salle de séminaires IRMA
Extreme meteorological events often occur in complex temporal configurations, where the impacts of one hazard may depend on the prior occurrence of others. Characterising such temporal dependencies is essential for understanding compound climate risks, yet remains challenging due to the discrete, heterogeneous, and clustered nature of extreme events. In this study, we apply temporal point process methods to characterise dependencies among extreme meteorological events occurring within appropriately defined spatial regions across Europe, focusing exclusively on their temporal structure.
We introduce an event-based framework in which extreme events are represented as marked temporal point processes, with marks describing key characteristics such as intensity or duration. Global first- and second-order temporal statistics are used to quantify clustering, co-occurrence, and directional dependencies between different types of extremes. In particular, we rely on directional cross-$K$ functions to assess whether the occurrence of one type of extreme event systematically modifies the short-term probability of subsequent events of another type.
Two complementary applications illustrate different facets of compound event analysis. First, we demonstrate the relevance of the framework for preconditioned compound events through a temporal analysis of wildfire-related meteorological extremes. Second, we examine temporal dependence between extreme precipitation, extreme wind, and extreme atmospheric instability across all European NUTS-2 regions.
Building on these second-order statistics, we develop formal tests of temporal independence to assess the significance of observed directional interactions between different types of extreme events. Overall, this temporal point process framework provides a rigorous and interpretable approach to the analysis of compound and preconditioned climate extremes, with direct applications to climate risk assessment and early-warning systems. -
Hugo Lebeau
Random Matrices and Tensors for Large-Dimensional Statistical Learning
17 avril 2026 - 11:00Salle de séminaires IRMA
This presentation will have two parts. The first part introduces an extension of spectral clustering on data streams. Assuming observations are made in an online fashion, we show how spectral clustering can be performed on the fly with a fixed amount of available memory. Studying this problem amounts to the spectral analysis of a particular Gram matrix in the high-dimensional regime, which we perform using tools from Random Matrix Theory. Based on our results, we describe the optimal memory policy and the corresponding clustering performance.
The second part of the presentation tackles the estimation of a planted low-rank signal in a large-dimensional tensor. This problem reveals a statistical-to-computational gap: a regime in which the maximum likelihood estimator is efficient, yet no polynomial-time algorithm can compute it. We study the performance of a procedure based on unfoldings, which is known to achieve the best algorithmic threshold, thereby revealing insights into the computational barrier.