Communication-efficient sparse regression: a one-shot approach

TL;DR

This paper introduces a one-shot distributed sparse regression method that averages debiased lasso estimators, achieving convergence rates comparable to the lasso in high-dimensional settings, and extends to generalized linear models.

Contribution

The paper presents a novel one-shot distributed regression technique that leverages averaging debiased estimators, reducing communication costs in high-dimensional data analysis.

Findings

Converges at the same rate as the lasso under certain conditions.

Effective in high-dimensional sparse regression.

Extended to generalized linear models.

Abstract

We devise a one-shot approach to distributed sparse regression in the high-dimensional setting. The key idea is to average "debiased" or "desparsified" lasso estimators. We show the approach converges at the same rate as the lasso as long as the dataset is not split across too many machines. We also extend the approach to generalized linear models.