Communication-efficient sparse regression

We devise a communication-efficient approach to distributed sparse regression in the high-dimensional setting. The key idea is to average "debiased" or "desparsified" lasso estimators. We show the ...