Robust Causal Discovery

Description du projet

Vers une découverte causale plus fiable

L’explosion des données au cours de la dernière décennie a favorisé les progrès dans des domaines tels que la science des données, les statistiques et l’économétrie. Alors que les méthodes traditionnelles se concentraient sur la recherche de modèles et d’associations, les chercheurs reconnaissent aujourd’hui l’importance des relations de cause à effet pour des informations plus approfondies. Cette évolution a entraîné une augmentation de la recherche sur l’inférence causale. Les méthodes actuelles supposent toutefois que les données sont propres et parfaitement structurées (une hypothèse souvent contredite par des erreurs de mesure et des anomalies). Soutenu par le programme Actions Marie Skłodowska-Curie, le projet ROCDISCO entend développer des méthodes de découverte causale robustes qui demeurent fiables malgré la contamination des données. En créant un cadre théorique, en concevant des techniques rigoureusement robustes et en testant des applications réelles, le projet renforcera la découverte causale et améliorera la fiabilité scientifique.

Objectif

RObust Causal DISCOvery

Due to technological advances, the available amount of data has increased tremendously over the last decade. The fields of data science, statistics, computer science and econometrics have followed this growth as they provide indispensable tools for translating data into insights and knowledge. Where data science was traditionally concerned with learning associations in data, it has recently become clear that causal relations often provide a deeper understanding and a stronger tool in many practical applications. This has led to the flourishing of causal inference with some of the most prestigious scientific awards going to pioneers in the field over the last decade.

“Can we learn causal mechanisms from observational data?” is one of the compelling questions that is occupying scientists all over the world. Where it was originally answered by skepticism, it has become clear that we are not completely powerless and there are indeed ways to infer causal structure from observational data under the right conditions. However, all of the current methods assume that the observed data perfectly follows the underlying causal structure. Unfortunately, real world data is often contaminated by anomalies and measurement errors, violating this assumption and thus weakening the reliability of methods for causal discovery.

This proposal aims to fill this gap by developing methods for causal discovery that remain efficient and reliable under data contamination. In particular, it (i) builds a theoretical framework for robust causal discovery, (ii) develops methods for causal discovery that are provably robust and correctly identify the causal structure and (iii) investigates the effect of contamination on real-world discovery tasks. As a result, in addition to advancing the theoretical understanding of causal discovery, this proposal builds a versatile toolbox to support scientists doing causal discovery and improve the reliability of their findings.