This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 6.896 Sublinear Time Algorithms November 02, 2004 Lecture 15 Lecturer: Eli BenSasson Scribe: Swastik Kopparty 1 Approximating Matrix Product in Sublinear Time Today we consider the problem of multiplying matrices in sublinear time. The results are from the work of Drineas, Kannan and Mahoney [1]. Given A R m n , B R n p , we wish to approximate A B R m p in sublinear time. Think of m, p n , so that the size of the output is sublinear in the size of the input. For starters, we need to define what we mean by approximating a matrix. will use the Frobenius norm as our measure of proximity. Definition 1 (Matrix Norms) For a m p matrix M , the Frobenius Norm of M is  M  F . = s X i [ m ] ,j [ p ] M 2 i,j . The L 2 norm of M is  M  2 = sup x R n  Mx  . In our context, we will say M approximates AB if  M AB  F is small. The Frobenius norm is related to the L 2 norm by the following inequalities:  M  2  M  F n  M  2 . Thus an approximation in Frobenius Norm gives some approximation in L 2 norm. We point out that [1] give more results on approximating the L 2 norm, which we wont describe. 1.1 The Sublinear Approximation Algorithm Key Observation The product AB , which is an m p matrix, is a sum of n rank one m p matrices....
View
Full
Document
 Spring '04
 RonittRubinfeld
 Algorithms

Click to edit the document details