Oskar Laverny

Oskar Laverny

Maître de Conférence

Aix-Marseille Université

Biography

Associate professor (Maître de conférence) at the Université Aix-Marseille (Marseille, FR), my research focus on high dimensional statistics and dependence structure estimations, with various fields of applications. Before that, I held postdoctoral positions at York University (Toronto, CA) and at the UCLouvain (Louvain-la-Neuve, BE), and before that, I did my PhD at Université Lyon 1 (Lyon, FR), under a CIFRE grant with SCOR SE. Long time ago, I graduated in actuarial sciences at ISFA (Lyon, FR).

Fully qualified actuary, I do have a taste for numerical code and open-source software, and all my work is freely available online. For examples in various formats, you may take a look at this paper about high dimensional Thorin measures, this Julia package about copulas, this blog post about latex automation, this deck of slides about version control for academics. Other interesting papers, slides, packages and more are available below.

Interests

  • High dimensional statistics
  • Dependence modeling
  • Code
  • Actuarial Sciences

Education

  • PhD: Dependence structure and risk agregation in high dimensions., 2019-2022

    ICJ & SCOR

  • Bachelor, Master and Diploma in Actuarial Sciences, 2015-2018

    ISFA

  • Bachelor in Mathematics, 2012-2015

    University of Straßbourg

Recent Publications

Local moment matching with Erlang mixtures under automatic roughness penalization

We consider the class of Erlang mixtures for the task of density estimation on the positive real line when the only available …

Parametric divisibility of stochastic losses

A probability distribution is n-divisible if its nth convolution root exists. While modeling the dependence structure between several …

Estimation of high dimensional Gamma convolutions through random projection

Multivariate generalized Gamma convolutions are distributions defined by a convolutional semi-parametric structure. Their flexible …

Estimation of multivariate generalized gamma convolutions through Laguerre expansions

The generalized gamma convolutions class of distributions appeared in Thorin’s work while looking for the infinite divisibility …

Empirical and non-parametric copula models with the cort R package

The R package cort implements object-oriented classes and methods to estimate, simulate and visualize certain types of non-parametric …

Dependence structure estimation using Copula Recursive Trees

We construct the Copula Recursive Tree (CORT) estimator: a flexible, consistent, piecewise linear estimator of a copula, leveraging …

Recent & Upcoming Talks

Shine of multiple dispatch: the `Copulas.jl` case.
Julia: the unique solution to an optimisation problem
Copulas.j: A fully Distribution.jl-complient copula package
Parametric divisibility of stochastic losses
Version control for academics
Estimation of High-dimensional Thorin measures
Estimation of High-dimensional Thorin measures
Estimation of High-dimensional Thorin measures
Estimation of multivariate generalized gamma convolutions
Julia, l'unique solution d'un problème d'optimisation
!Canceled! Copula estimation via machine learning
Cort: The Copula Recursive Tree

Teaching

 
 
 
 
 

Inférence statistique des tests - CM + TD

ENS Lyon

Sep 2021 – Jan 2022
  • 36h integrated lectures (main lecture + tutorials) given to third year students in Economics.
  • Content: From the definition of a probability to mean and variance tests in the Gaussian or asymptotically Gaussian cases, chi-squared tests, with a particular emphasis on Cochran’s proof. Kolomogorov-Smirnov test with a full proof. Maximum Likelyhood estimation and likelyhood-ratio tests.
  • The particularity was that I had the opportunity to create and design the content and the form of lectures myself.
 
 
 
 
 

Probabilité, combinatoire et statistiques - TD

University Claude Bernard Lyon 1

Jan 2021 – May 2021
  • 30h tutorials given to third year students in Mathematics.
  • Content: From the definition of a probability to mean and variance tests in the Gaussian or asymptotically Gaussian cases, chi-squared tests. Discrete time markov chains. Numerical applications in R
 
 
 
 
 

Statistiques pour l’informatique - TD

University Claude Bernard Lyon 1

Sep 2020 – Jan 2022
  • 30h tutorials given to second year students in Informatics, in Falls 2020 and Falls 2021.
  • Content: From the definition of a probability to mean and variance tests in the Gaussian or asymptotically Gaussian cases, chi-squared tests. Numerical applications in R.

Recent Posts

Version control for academics

A few months ago, I was proposed the opportunity to give a talk at ISBA’s Young researcher’s day 2023 (UcLouvain, Louvain-la-neuve, BE) about version control. I already wrote a bit on git and GitHub for latex writing and even more automation.

Estimation and sampling of copulas in Julia with Copulas.jl

Announce I am proud to annonce the publication and the registration of my new Julia package, Copulas.jl. As it’s name suggests, Copulas.jl is a package that implements methods and tools to work with an arround copulas in the Julia programming language.

Automatic latex resume with github action and gitinfo2 watermark

The problematic A latex-written resume is always a nice thing to have: easy to update, practical to integrate .bib bibliographies, and automatic management of the look (you only provide its content).

Experiences

 
 
 
 
 

Associate Professor (Maître de conférence)

Aix-Marseille Université – SESSTIM

Sep 2023 – Present Marseille
 
 
 
 
 

Post-Doctoral Researcher

UCLouvain - ISBA

Oct 2022 – Aug 2023 Paris & Lyon
PostDoc funded by the FNRS.
 
 
 
 
 

Post-Doctoral Researcher

York university – RISC

Jul 2022 – Sep 2023 Toronto
PostDoc funded by the Fields institute.
 
 
 
 
 

PhD

SCOR & Univ Lyon 1

Jun 2019 – Jun 2022 Paris & Lyon
The title of the PhD is “Dependence structures and risk agregations in high dimensions.” The PhD was defended on May, the 30th of 2022.
 
 
 
 
 

Actuary

L’Auxiliaire

May 2017 – Sep 2018 Lyon
Non Life reserving of french builder’s insurence in a Solvency II context :

  • Analysis
  • Construction of an USP
  • Redaction of an actuarial thesis on it.