Sparsifying generalized linear models
Abstract
We consider the sparsification of sums where for vectors and functions . We show that -approximate sparsifiers of with support size exist whenever the functions are symmetric, monotone, and satisfy natural growth bounds. Additionally, we give efficient algorithms to compute such a sparsifier assuming each can be evaluated efficiently. Our results generalize the classic case of sparsification, where , for , and give the first near-linear size sparsifiers in the well-studied setting of the Huber loss function and its generalizations, e.g., for . Our sparsification algorithm can be applied to give near-optimal reductions for optimizing a variety of generalized linear models including regression for to high accuracy, via solving sparse regression instances with , plus runtime proportional to the number of nonzero entries in the vectors .
Keywords
Cite
@article{arxiv.2311.18145,
title = {Sparsifying generalized linear models},
author = {Arun Jambulapati and James R. Lee and Yang P. Liu and Aaron Sidford},
journal= {arXiv preprint arXiv:2311.18145},
year = {2023}
}