Research output from my time at Safety Line (2011–2020) and LumenAI (2020-2022), in collaboration with INRIA and ENAC.
In Actes de la conférence CAID 2021 (Conference on Artificial Intelligence for Defense) — 2021
Online graph clustering approach for detecting suspicious activities in a computing system based on the detection of communities in a network.
Download / ViewIn ESAIM Probability & Statistics, 2021 — 2021
Quantify the closeness of a newly observed curve to a given sample of random functions, supposed to have been sampled from the same distribution. Application to a dataset of real aircraft trajectories.
Co-authored during PhD co-supervision with INRIA (2015–2018).
Download / ViewPatent — 2020
Uses flight data and weather information to optimize flight plans for a fleet of aircraft by determining the best route considering fuel consumption and meteorological conditions.
Download / ViewPatent — 2019
Method for analysing an aircraft flight to assess exposure to a given risk. Filed during the Safety Line period, arising from R&D work on automated flight risk scoring using large-scale flight data recorder processing and machine learning.
Download / ViewIn Journal of Guidance, Control, and Dynamics. — 2019
Trajectory optimisation under distributional constraints using Gaussian mixture penalties. Applied to aircraft climb profile optimisation to balance fuel efficiency with operational constraints.
Co-authored during PhD co-supervision with INRIA (2015–2018).
Download / ViewIn 7th European Conference for Aeronautics ans Space Science, EUCASS — 2017
Four new Maximum Likelihood based approaches for aircraft dynamics identification are presented and compared. The motivation is the need of accurate dynamic models for minimizing aircraft fuel consumption using optimal control techniques. A robust method for building aerodynamic models is also suggested. All these approaches were validated using real flight data from 25 different aircraft
Co-authored during PhD co-supervision with INRIA (2015–2018).
Download / ViewPreprint — 2017
Clustering of functional data (curves) using second-order statistics. Applied to runway condition monitoring from flight data recorder streams — detecting degraded surface conditions from aircraft behaviour during landing roll.
Collaboration with Ecole Nationale de l’Aviation Civile (ENAC)
Download / ViewPreprint — 2017
Identification of structured dynamical systems from large-scale flight data using block sparse linear modelling. Application to aircraft trajectory analysis and system identification from flight data recorder streams.
Co-authored during PhD co-supervision with INRIA (2015–2018).
Download / ViewStatistics and Computing — 2016
Theoretical and empirical analysis of variable importance measures in random forests in the presence of correlated predictors. Studies the bias introduced by correlation on standard importance scores and proposes corrections.
Core research from my PhD at Université Pierre et Marie Curie.
Download / ViewUniversité Pierre et Marie Curie — 2015
L’objectif de cette thèse est de proposer un ensemble d’outils méthodologiques pour répondre à la problématique de l’analyse des données de vol. Les travaux présentés dans ce manuscrit s’articulent autour de deux thèmes statistiques : la sélection de variables en apprentissage supervisé d’une part et l’analyse des données fonctionnelles d’autre part. Nous présentons diverses contributions liées à la mesure d’importance par permutation de l’algorithmes des forêts aléatoires.
Manuscipt of my PhD at Université Pierre et Marie Curie.
Download / ViewJournal of Computational Statistics & Data Analysis — 2015
Extension of variable importance measures to grouped variables in random forests. Application to multivariate functional data from flight data recorders, where variables are naturally grouped by physical sensor type. Enables identification of the most relevant sensor groups for a given prediction task.
Core research from my PhD at Université Pierre et Marie Curie.
Download / View