`

Timezone: »

 
Electromagnetic neural source imaging under sparsity constraints with SURE-based hyperparameter tuning
Pierre-Antoine Bannier · Quentin Bertrand · Joseph Salmon · Alexandre Gramfort

Estimators based on non-convex sparsity-promoting penalties were shown to yield state-of-the-art solutions to the magneto-/electroencephalography (M/EEG) brain source localization problem. In this paper we tackle the model selection problem of these estimators: we propose to use a proxy of the Stein's Unbiased Risk Estimator (SURE) to automatically select their regularization parameters. The effectiveness of the method is demonstrated on realistic simulations and 30 subjects from the Cam-CAN dataset. To our knowledge, this is the first time that sparsity promoting estimators are automatically calibrated at such a scale. Results show that the proposed SURE approach outperforms cross-validation strategies and state-of-the-art Bayesian methods both computationally and statistically.

Author Information

Pierre-Antoine Bannier (INRIA)
Quentin Bertrand (Mila)
Joseph Salmon (Université de Montpellier)
Alexandre Gramfort (INRIA)

More from the Same Authors