Random sampling for estimating the performance of fast summations

Thumbnail Image

Files

CS-TR-4969.pdf (56.99 KB)
No. of downloads: 620

Publication or External Link

Date

2010-10-18

Advisor

Citation

DRUM DOI

Abstract

Summation of functions of N source points evaluated at M target points occurs commonly in many applications. To scale these approaches for large datasets, many fast algorithms have been proposed. In this technical report, we propose a Chernoff bound based efficient approach to test the performance of a fast summation algorithms providing a probabilistic accuracy. We further validate and use our approach in separate comparisons.

Notes

Rights