Social-sparsity brain decoders: faster spatial sparsity

TL;DR

This paper introduces social-sparsity, a computationally efficient brain decoding method that achieves comparable accuracy to total variation models and better interpretability than graph-net by focusing on local voxel neighborhoods.

Contribution

The paper proposes social-sparsity, a new structured shrinkage operator for brain decoding that reduces computational costs while maintaining high accuracy and interpretability.

Findings

Social-sparsity performs nearly as well as total-variation models.

It outperforms graph-net in classification tasks.

It clearly highlights predictive brain regions.

Abstract

Spatially-sparse predictors are good models for brain decoding: they give accurate predictions and their weight maps are interpretable as they focus on a small number of regions. However, the state of the art, based on total variation or graph-net, is computationally costly. Here we introduce sparsity in the local neighborhood of each voxel with social-sparsity, a structured shrinkage operator. We find that, on brain imaging classification problems, social-sparsity performs almost as well as total-variation models and better than graph-net, for a fraction of the computational cost. It also very clearly outlines predictive regions. We give details of the model and the algorithm.