Proof that Finite-Dimensional Normed Vector Spaces are Closed, Assignments of Theories of Communication

The solution to problem 1 of homework assignment 4 in ece-542 digital communication theory, spring 2004. The problem asks to prove by induction that any finite-dimensional normed vector space is closed. The solution is mathematically precise and justifies each step of the proof.

Typology: Assignments

Pre 2010

Uploaded on 08/18/2009

koofers-user-pm1
koofers-user-pm1 🇺🇸

10 documents

1 / 1

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
ECE-542 Digital Communication Theory
Spring 2004
Homework Assignment 4 solutions
Problem 1. Reconstruct the proof done in class which shows that any finite-dimensional normed
vector space is closed. Be precise, mathematically descriptive of the concepts, and justify your
steps thoroughly along the way.
We prove this by induction on the dimension. First note that by a norm, || · ||,wemeana
mapping from the vector space to the reals such that (1) ||u||≥0 for any vector uand ||u|| =0
only if u=0;(2)||αu|| =|α|||u||, for any scalar αand vector u; and (3) ||u+v|| ||u|| +||v|| for
any vectors uand v.
Induction initialization:
Assume that the dimension is 1. We need to show that the one-dimensional space, say M1,is
closed. That is, if limn→∞ ||xxn|| =0andxnM1, then xM1. Note that we can write
xn=anv1, where v1spans M1and anis a scalar (real or complex). We will show that the sequence
anis bounded, which means that it has a convergent subsequence ani. (This is a consequence of the
completeness of the real/comples field, which is called the Bolzano Wierstrass Theorem). Assume
therefore that ania0, for some scalar a0. It follows that limi→∞ ||xnia0v1|| = 0. But since
xnconverges to x, any subsequence of it also converges to x. Thus, limi→∞ ||xxni|| = 0. Hence
x=a0v1M1. It remains to prove that anis bounded. Note that ||xn|| ||xxn|| +||x|| <,
and hence |an|=||xn||/||v1|| <.
Induction hypothesis:
Any k-dimensional space Mk,k1, is closed.
Proof of the induction step:
We show that any (k+ 1)-dimensional space Mk+1 , is closed. That is, if xnMk+1 is conver-
gent to x, then xMk+1. We can write xn=a1
1u1+...+ak+1
nuk+1, where u1,...,u
k+1 span
Mk+1. We now show that the sequence ak+1
nis bounded. We prove this by contradiction: Suppose
that ak+1
nis unbounded, and without loss of generality, assume that ak+1
n→∞. Observe that
xn/ak+1
n=a1
nu1/ak+1
n+...+ak
nu1/ak+1
n+uk+1.Now||xn/ak+1
n|| 0 since ak+1
n→∞.Thus,
a1
nu1/ak+1
n+...+ak
nuk/ak+1
n→−uk+1. But since a1
nu1/ak+1
n+...+ak
nuk/ak+1
nis in a k-dimensional
space spanned by u1,...,u
k, which is closed (by the induction hypothesis), its limit must be in
the space spanned by u1,...,u
kand not uk+1, which is a contradiction. Hence, ak+1
nmust be
bounded. Consequently, it has a convergent subsequence ak+1
ni, with a limit say ak+1
0. Observe that
a1
niu1+...+ak
niuk=xniak+1
niuk+1, which converges to xak+1
0uk+1.Buta1
niu1+...+ak
niuk
is in a k-dimensional space spanned by u1,...,u
k, its limit, which is xak+1
0uk+1,mustalsobe
in the same space, namely, xak+1
0uk+1 =b1u1+...+bkuk, for some scalars b1,...,b
k+1.This
implies that x=b1u1+...+bkuk+ak+1
0uk+1.Thus,xMk+1 as desired.
From the text: 3.29, 3.32, 3.33, 3.38 (pick up solution from my office).
1

Partial preview of the text

Download Proof that Finite-Dimensional Normed Vector Spaces are Closed and more Assignments Theories of Communication in PDF only on Docsity!

ECE-542 Digital Communication Theory Spring 2004 Homework Assignment 4 solutions

Problem 1. Reconstruct the proof done in class which shows that any finite-dimensional normed vector space is closed. Be precise, mathematically descriptive of the concepts, and justify your steps thoroughly along the way.

We prove this by induction on the dimension. First note that by a norm, || · ||, we mean a mapping from the vector space to the reals such that (1) ||u|| ≥ 0 for any vector u and ||u|| = 0 only if u = 0; (2) ||αu|| = |α|||u||, for any scalar α and vector u; and (3) ||u + v|| ≤ ||u|| + ||v|| for any vectors u and v.

Induction initialization: Assume that the dimension is 1. We need to show that the one-dimensional space, say M 1 , is closed. That is, if limn→∞ ||x − xn|| = 0 and xn ∈ M 1 , then x ∈ M 1. Note that we can write xn = anv 1 , where v 1 spans M 1 and an is a scalar (real or complex). We will show that the sequence an is bounded, which means that it has a convergent subsequence ani. (This is a consequence of the completeness of the real/comples field, which is called the Bolzano Wierstrass Theorem). Assume therefore that ani → a 0 , for some scalar a 0. It follows that limi→∞ ||xni − a 0 v 1 || = 0. But since xn converges to x, any subsequence of it also converges to x. Thus, limi→∞ ||x − xni || = 0. Hence x = a 0 v 1 ∈ M 1. It remains to prove that an is bounded. Note that ||xn|| ≤ ||x − xn|| + ||x|| < ∞, and hence |an| = ||xn||/||v 1 || < ∞.

Induction hypothesis: Any k-dimensional space Mk, k ≥ 1, is closed.

Proof of the induction step: We show that any (k + 1)-dimensional space Mk+1, is closed. That is, if xn ∈ Mk+1 is conver- gent to x, then x ∈ Mk+1. We can write xn = a^11 u 1 +... + ak n+1 uk+1, where u 1 ,... , uk+1 span Mk+1. We now show that the sequence ak n+1 is bounded. We prove this by contradiction: Suppose that ak n+1 is unbounded, and without loss of generality, assume that ak n+1 → ∞. Observe that xn/ak n+1 = a^1 nu 1 /ak n+1 +... + aknu 1 /ak n+1 + uk+1. Now ||xn/ak n+1 || → 0 since ak n+1 → ∞. Thus, a^1 nu 1 /ak n+1 +.. .+aknuk/ak n+1 → −uk+1. But since a^1 nu 1 /ak n+1 +.. .+aknuk/ak n+1 is in a k-dimensional space spanned by u 1 ,... , uk, which is closed (by the induction hypothesis), its limit must be in the space spanned by u 1 ,... , uk and not uk+1, which is a contradiction. Hence, ak n+1 must be bounded. Consequently, it has a convergent subsequence ak n+1i , with a limit say ak 0 +1. Observe that a^1 ni u 1 +... + akni uk = xni − ak n+1i uk+1, which converges to x − ak 0 +1 uk+1. But a^1 ni u 1 +... + akni uk is in a k-dimensional space spanned by u 1 ,... , uk, its limit, which is x − ak 0 +1 uk+1, must also be in the same space, namely, x − ak 0 +1 uk+1 = b 1 u 1 +... + bkuk, for some scalars b 1 ,... , bk+1. This implies that x = b 1 u 1 +... + bkuk + ak 0 +1 uk+1. Thus, x ∈ Mk+1 as desired.

From the text: 3.29, 3.32, 3.33, 3.38 (pick up solution from my office).