A Fast Randomized Algorithm for Overdetermined Linear Least-Squares Regression
Abstract
We introduce a randomized algorithm for overdetermined linear least-squares regression. Given an arbitrary full-rank m x n matrix A with m greater than or equal to n, any m x 1 vector b, and any positive real number epsilon, the procedure computes an n x 1 vector x which minimizes the spectral norm ||A(sub x)- b|| to relative precision epsilon. The algorithm typically requires O ((log(n) + log(1/epsilon))mn+n(exp 3)) floating-point operations. This cost is less than the O(mn(exp 2)) required by the classical schemes based on QR-decompositions or bidiagonalization. We present several numerical examples illustrating the performance of the algorithm.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 28, 2008
- Accession Number
- ADA482134
Entities
People
- Mark Tygert
- Vladimir Rokhlin, Jr.
Organizations
- Yale University