inria_bilevel_optim's Introduction

Bilevel optimization for structured sparse models

This is a working repository for my internship at INRIA with Alexandre Gramfort, Joseph Salmon and Quentin Bertrand. You can find an implementation of MCFD SURE in the mtl package, as well as experiment results made on CamCan dataset for the neuroimaging source localization inverse problem.

This internship has mainly resulted in the publication of a paper (LINK TO BE DISCLOSED) and in the implementation of SURE in MNE package (https://github.com/mne-tools/mne-python).

inria_bilevel_optim's People

Contributors

Stargazers

Watchers

inria_bilevel_optim's Issues

TODO tex

ne pas mettre de cite:
utilise citep quand "tu peux enlever la citation de la phrase", "SURE was applied on MEEG data citep{Bannier2021}"
utilise citet quand "tu ne peux pas enlever la citation de la phrase" "\citet{Bannier2021} applied SURE on MEEG data"
bien séparer 1 idée par paragraphe
" \cite{Pedregosa16} has showed that cross-validating the hyperparameter on held-out sets is not a suitable approach for inverse problems" je crois que ce n'est pas lui, si tu regardes la reviex de Arlot et Célisse sur la CV il y a des réf là dessus
paragraphe setting d'hyperparam: bien séparer en 2 catégories: Bayesian (qui se subdvise en plusieurs catéories) et hyperparameter optimization (CV, SURE etc)
pas de newline
\textit{\textbf{Notation}} >> just \textbf ou \paragraph
il faut que tu utilises l'option remove trailing spaces on save de vscode: https://stackoverflow.com/questions/30884131/remove-trailing-spaces-automatically-or-with-a-shortcut#
commencer la description des expes

Recommend Projects