Approche basée sur l'apprentissage automatique pour la détection d'anomalies lors de la surveillance d'ouvrages
International audience
Exploring Risk Factors and Prediction Models for Sudden Cardiac Death with Machine Learning
Sudden cardiac death (SCD) is defined as a sudden natural death presumed to be of cardiac cause, heralded by abrupt loss of consciousness in the presence of witness, or in the absence of witness occurring within an hour after the onset of symptoms. Despite progress in clinical profiling and interventions, it remains a major public health problem, a...
Contributions to a Theory of Pure Exploration in Sequential Statistics
This thesis lies in the fields of artificial intelligence, sequential statistics and optimization. We focus on the problem of best (in expectation) arm identification in unstructured muti-armed bandits. This problem has two approaches with very different levels of understanding. The fixed-confidence framework is the best understood: asymptotically ...
Apports de la statistique bayésienne pour l'estimation du risque de cancer radio-induit après exposition aux examens sca...
Ces dernières décennies, une multiplication des examens diagnostiques utilisant les rayonnements ionisants (RIs), comme les examens scanners, a été observée. Cependant, les effets sanitaires à long terme de ces expositions radiologiques sont peu connus. La cohorte Enfant Scanner a ainsi été mise en place en 2009 à l'IRSN afin d'étudier les risques ...
Modélisation du biais dans les recrutements : étude de l'influence d'un biais dans les données d'apprentissage de différ...
Grâce aux récentes avancées de l'intelligence artificielle, de nouvelles méthodes automatiques pour l'aide à la décision sont mises au service du public (comme la reconnaissance d'images par exemple). Dans la science du recrutement, des solutions automatiques ont également été développées et sont vendues par des start-up. Or, à cause du secret indu...
Une méthode d'approximation des effets de Shapley en grande dimension
En apprentissage statistique supervisé et en analyse d'incertitudes de codes de calcul, les mesures d'importance (ou indices de sensibilité) ont pour but de quantifier l'influence des entrées (ou covariables) du modèle d'apprentissage ou du code de calcul sur sa sortie. Du fait de leur simplicité d'interprétation même en présence de fortes dépendan...
Linear Dimensionality Reduction
These notes are an overview of some classical linear methods in Multivariate Data Analysis. This is a good old domain, well established since the 60's, and refreshed timely as a key step in statistical learning. It can be presented as part of statistical learning, or as dimensionality reduction with a geometric flavor. Both approaches are tightly l...
Optimal transport for graph representation learning
A key challenge in Machine Learning (ML) is to design models able to learn efficiently from graphs, characterized by nodes with attributes and a prescribed structure encoding their relationships. Graph Representation Learning (GRL) aims to encode these two sources of heterogeneity into a vectorial graph embedding easing downstream tasks. In this fi...
Machine learning for genomics and imaging data integration applied to neuro-oncology
Diffuse Intrinsic Pontine Glioma (DIPG) is a rare brain tumour located in the pons, mainly seen in children aged 5 to 7 years. It is considered one of the most aggressive paediatric tumours, with a survival rate of less than 10% beyond two years after diagnosis and a median overall survival of less than one year. DIPG is classified as a diffuse mid...
Understanding Post-hoc Explainers: The Case of Anchors
In many scenarios, the interpretability of machine learning models is a highly required but difficult task. To explain the individual predictions of such models, local model-agnostic approaches have been proposed. However, the process generating the explanations can be, for a user, as mysterious as the prediction to be explained. Furthermore, inter...
