What question did this study set out to answer?

This research aims to develop a generalised information criteria (gic) to improve model selection for sparse models.

May 22, 2026Open Access

A note on generalised information criteria for structured sparse models

Key Points

This research aims to develop a generalised information criteria (gic) to improve model selection for sparse models.
Introduced a generalised information criteria that considers sparsity patterns.
Derived asymptotic and nonasymptotic results for effective model selection.
Illustrated application of gic in group LASSO and low-rank matrix regression contexts.
The gic provides a robust framework for selecting regularisation parameters in high-dimensional settings.
Examples demonstrated the efficacy of gic in managing sparsity in model selection.

Abstract

Summary We propose a generalised information criteria ( gic ) that accounts for sparsity pattern in the model. We obtain both asymptotic and nonasymptotic results for model selection. Moreover, we show that the gic is useful for selecting the regularisation parameter in regularised estimation in high‐dimensional scenarios. The results are illustrated in two examples: group LASSO in the context of generalised linear regressions and low‐rank matrix regression.

Ask AI

Helpful

Bookmark

View Full Paper