# Fact 27 vector p norms let x cn with elements x x1 p

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: x for all α ∈ C, x ∈ Cn . The vector p-norms below are useful for computational purposes, as well as analysis. Fact 2.7 (Vector p -Norms). Let x ∈ Cn with elements x = x1 p-norm 1/p ... xn T . The n x p = is a vector norm. j =1 |xj |p p ≥ 1, , Example. • If ej is a canonical vector, then ej • If e = 1 e 1 1 = n, T ∈ Rn , then e ··· 1 ∞ = 1, p = 1 for p ≥ 1. e p = n1/p , 1 &lt; p &lt; ∞. The three p-norms below are the most popular, because they are easy to compute. • One norm: x 1 n j =1 |xj |. = • Two (or Euclidean) norm: x 2 = • Inﬁnity (or maximum) norm: x Example. If x = 1 x 1 2 n ··· 1 = n(n + 1), 2 x T 2 = ∞ n 2 j =1 |xj | √ = x∗x. = max1≤j ≤n |xj |. ∈ Rn , then 1 n(n + 1)(2n + 1), 6 x ∞ = n. The inequalities below bound inner products in terms of norms. Fact 2.8. Let x , y ∈ Cn . Then Hölder inequality: |x ∗ y | ≤ x 1 y ∞ Cauchy–Schwarz inequality: |x ∗ y | ≤ x Moreover, |x ∗ y | = x 2 y 2 2 y 2. if and only if x and y are multiples of each other. Example. Let x ∈ Cn with elements x = x1 · · · xn and Cauchy–Schwarz inequality imply, respectively, n n i =1 xi ≤ n max |xi |, 1≤i ≤n i =1 xi ≤ T √ . The Hölder inequality n x 2. Downloaded 01/17/14 to 143.215.200.123. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php 2.5. Vector Norms 31 Deﬁnition 2.9. A nonzero vector x ∈ Cn is called unit-norm vector in the · norm if x = 1. The vector x/ x has unit norm. Example. Let e be the n × 1 vector of all ones. Then 1= e ∞ = 1 e n 1 = √e n 1 . 2 The canonical vectors ei have unit norm in any p-norm. Normwise Errors. We determine how much information the norm of an error gives about individual, componentwise errors. Deﬁnition 2.10. If x is an approximation to a vector x ∈ Cn , then x − x is a ˜ ˜ x −x ˜ x −x ˜ normwise absolute error. If x = 0 or x = 0, then x and x are normwise ˜ ˜ relative errors. How much do we lose when we replace componentwise errors by normwise errors? For vectors x , x ∈ Cn , the inﬁnity norm is equal...
View Full Document

Ask a homework question - tutors are online