OverSketch: Approximate Matrix Multiplication for the Cloud
Distributed, Parallel, and Cluster Computing
2019-02-25 v2 Information Theory
math.IT
Abstract
We propose OverSketch, an approximate algorithm for distributed matrix multiplication in serverless computing. OverSketch leverages ideas from matrix sketching and high-performance computing to enable cost-efficient multiplication that is resilient to faults and straggling nodes pervasive in low-cost serverless architectures. We establish statistical guarantees on the accuracy of OverSketch and empirically validate our results by solving a large-scale linear program using interior-point methods and demonstrate a 34% reduction in compute time on AWS Lambda.
Cite
@article{arxiv.1811.02653,
title = {OverSketch: Approximate Matrix Multiplication for the Cloud},
author = {Vipul Gupta and Shusen Wang and Thomas Courtade and Kannan Ramchandran},
journal= {arXiv preprint arXiv:1811.02653},
year = {2019}
}
Comments
Published in Proc. IEEE Big Data 2018. Updated version provides details of distributed sketching and highlights other advantages of OverSketch