# E i i 11 17 singular values the singular values of a

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: σ y with σ = A 2 . There exists V2 ∈ IRn×(n−1) and U2 ∈ IRm×(m−1) so V = [x V2 ] ∈ IRn×n and U = [y U2 ] ∈ IRm×m are orthogonal. It follows that σw σ y AV2 = U AV = U ≡ A1 0B for some w and B . Since A1 σ w 2 ≥ (σ 2 + w w)2 2 we have A1 2 ≥ (σ 2 + w w). But σ 2 = A 2 be 0. By induction, we complete this proof. 2 2 = A1 2 , and thus w must 2 10 / 17 Singular vectors As A = U ΣV , by comparing the columns in the equations AV = U Σ and A U = V Σ , it is easy to show Avi A ui = σ ui = σ vi where i = 1, . . . , min(m, n) The σi are the singular values of A and the vectors ui and vi are the i -th left singular vector and i -th right singular vector respectively U is a set of eigenvectors of AA ∈ IRm×m Σ is a diagonal matrix whose values are the square root of eigenvalues of AA ∈ IRm×m V is a set of eigenvectors of A A ∈ IRn×n It can be shown that singular values σi are the square roots of √ eigenvalues, λi , i.e., σi = λi 11 / 17 Singular values The singular values of a matrix A are precisely the lengths of the semi-axes of the hyperellipsoid E deﬁned by E = {Ax : x 2 = 1} The semi-axes are described by the singular vectors The SVD reveals the struct...
View Full Document

## This document was uploaded on 02/10/2014.

Ask a homework question - tutors are online