Distributed Least Squares in Small Space via Sketching and Bias Reduction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Garg, Sachin, Tan, Kevin, Dereziński, Michał
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916239442444288
author Garg, Sachin
Tan, Kevin
Dereziński, Michał
author_facet Garg, Sachin
Tan, Kevin
Dereziński, Michał
contents Matrix sketching is a powerful tool for reducing the size of large data matrices. Yet there are fundamental limitations to this size reduction when we want to recover an accurate estimator for a task such as least square regression. We show that these limitations can be circumvented in the distributed setting by designing sketching methods that minimize the bias of the estimator, rather than its error. In particular, we give a sparse sketching method running in optimal space and current matrix multiplication time, which recovers a nearly-unbiased least squares estimator using two passes over the data. This leads to new communication-efficient distributed averaging algorithms for least squares and related tasks, which directly improve on several prior approaches. Our key novelty is a new bias analysis for sketched least squares, giving a sharp characterization of its dependence on the sketch sparsity. The techniques include new higher-moment restricted Bai-Silverstein inequalities, which are of independent interest to the non-asymptotic analysis of deterministic equivalents for random matrices that arise from sketching.
format Preprint
id arxiv_https___arxiv_org_abs_2405_05343
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Distributed Least Squares in Small Space via Sketching and Bias Reduction
Garg, Sachin
Tan, Kevin
Dereziński, Michał
Data Structures and Algorithms
Machine Learning
Numerical Analysis
Matrix sketching is a powerful tool for reducing the size of large data matrices. Yet there are fundamental limitations to this size reduction when we want to recover an accurate estimator for a task such as least square regression. We show that these limitations can be circumvented in the distributed setting by designing sketching methods that minimize the bias of the estimator, rather than its error. In particular, we give a sparse sketching method running in optimal space and current matrix multiplication time, which recovers a nearly-unbiased least squares estimator using two passes over the data. This leads to new communication-efficient distributed averaging algorithms for least squares and related tasks, which directly improve on several prior approaches. Our key novelty is a new bias analysis for sketched least squares, giving a sharp characterization of its dependence on the sketch sparsity. The techniques include new higher-moment restricted Bai-Silverstein inequalities, which are of independent interest to the non-asymptotic analysis of deterministic equivalents for random matrices that arise from sketching.
title Distributed Least Squares in Small Space via Sketching and Bias Reduction
topic Data Structures and Algorithms
Machine Learning
Numerical Analysis
url https://arxiv.org/abs/2405.05343