Transfer Learning for Penalized Gaussian Graphical Models

Details

Supervisors

Eugen Pircalabelu

Faculty

Faculté des sciences

Degree label

Master [120] en statistique, orientation biostatistiques, à finalité spécialisée

Abstract

Penalized Gaussian Graphical Models aim to describe the graphical structure underlying the variables. Transfer Learning allows to add one or more auxiliary studies as additional information for the model estimation, which could then lead to enhanced performance of the estimator. In this thesis, we aim to explain the theory behind Transfer Learning applied to Gaussian Graphical Models and compare the performance of the corresponding estimator, Trans-CLIME, against two well-known competitors, the CLIME and Glasso estimators.