Proofs about invertible linear functions

In summary: Yes, you are correct. We cannot say for sure that $||\sum_{k=0}^{\infty }(I-A_p)^k|| \to \frac{1}{1-||I-A_p||}$, but we can say that it is bounded by it. Then we can use the fact that $||\sum_{k=0}^{\infty }(I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)^k||$ and use the fact that the sum of a geometric series converges to $\frac{1}{1-||I-A_p||}$ to show that $||\sum_{k=0}
  • #1
i_a_n
83
0
Let $G\subset L(\mathbb{R}^n;\mathbb{R}^n)$ be the subset of invertible linear transformations.

a) For $H\in L(\mathbb{R}^n;\mathbb{R}^n)$, prove that if $||H||<1$, then the partial sum $L_n=\sum_{k=0}^{n}H^k$ converges to a limit $L$ and $||L||\leq\frac{1}{1-||H||}$.

b) If $A\in L(\mathbb{R}^n;\mathbb{R}^n)$ satisfies $||A-I||<1$, then A is invertible and $A^{-1}=\sum_{k=0}^{\infty }H^k$ where $I-A=H$. (Hint: Show that $AL_n=H^{n+1}$)

c) Let $\varphi :G\rightarrow G$ be the inversion map $\varphi(A)=A^{-1}$. Prove that $\varphi$ is continuous at the identity I, using the previous two facts.

d) Let $A, C \in G$ and $B=A^{-1}$. We can write $C=A-K$ and $\varphi(A-K)=c^{-1}=A^{-1}(I-H)^{-1}$ where $H=BK$. Use this to prove that $\varphi$ is continuous at $A$.
I have little ideas about these questions. What's your answers? Thank you!
 
Last edited:
Physics news on Phys.org
  • #2
a) Choose a submultiplicate norm so that
||AB|| $\leq$ ||A||||B||.
Now the sequence of partial sums $L_n$ converges to a limit L, then
$lim_{n \to \infty} ||L_n - L || = 0$,
Now $L_n = I + L + ... + L^{n}$. so $||L_n|| = ||I+...+H^n|| \leq ||I|| + ||H|| + ... + ||H^n|| \leq ||I|| + ||H|| + ||H||||H||+ ... + ||H||^n$ <-- by the submultiplicate norm.

so since ||H|| $< $ 1, then ||L_n - L|| $\leq 1 + ||H|| + ... + ||H||^n$, then assume $||H|| = p$, then this is just a geometric series, $1+|p|+...+|p|^n$, and this converges for $|p| < 1$, which is true. and $||L_n - L|| \leq 1 + |p| +...+|p|^n \leq \frac{1}{1-|p|}$ or $\frac{1}{1-||H||}$

b)if ||A-I|| < I
note that $(A-I)*(-I-A-A^2-...) = I $
Now we only need to show that $(-I-A-A^2...)$ converges.I have almost b,c done but i need to think about some details.
 
  • #3
jakncoke said:
a) Choose a submultiplicate norm so that
||AB|| $\leq$ ||A||||B||.
Now the sequence of partial sums $L_n$ converges to a limit L, then
$lim_{n \to \infty} ||L_n - L || = 0$,
Now $L_n = I + L + ... + L^{n}$. so $||L_n|| = ||I+...+H^n|| \leq ||I|| + ||H|| + ... + ||H^n|| \leq ||I|| + ||H|| + ||H||||H||+ ... + ||H||^n$ <-- by the submultiplicate norm.

so since ||H|| $< $ 1, then ||L_n - L|| $\leq 1 + ||H|| + ... + ||H||^n$, then assume $||H|| = p$, then this is just a geometric series, $1+|p|+...+|p|^n$, and this converges for $|p| < 1$, which is true. and $||L_n - L|| \leq 1 + |p| +...+|p|^n \leq \frac{1}{1-|p|}$ or $\frac{1}{1-||H||}$

b)if ||A-I|| < I
note that $(A-I)*(-I-A-A^2-...) = I $
Now we only need to show that $(-I-A-A^2...)$ converges.I have almost b,c done but i need to think about some details.

Now I especially need proofs of c and d...I can prove a and b.

What $\varphi$ is continuous at I and A means? What is needed to prove?
 
  • #4
You need to prove that as $A \to I$, then $\phi(A) \to \phi(I) = I$

or $||\phi(A) - I|| \to 0$ for any sequence of Matricies $A_n \to I$.
 
  • #5
jakncoke said:
You need to prove that as $A \to I$, then $\phi(A) \to \phi(I) = I$

or $||\phi(A) - I|| \to 0$ for any sequence of Matricies $A_n \to I$.

Can you show me how to prove c) and d)? Thank you a lot!
 
  • #6
You can see the truth of c) by this

So for any sequence $A_n \to I$, we can always pick a $N \in \mathbb{N}$ such that $\forall p \geq N$, $||A_p - I||< 1$

In b), we proved that $A_p$ has an inverse. $A_p^{-1]}$ or the form
$A_p^{-1} = \sum_{k=0}^{\infty} (I-A_p)^k$.

Expanding it out a bit $I + (I - A_p) + (I -A_p)^2...$ since $||\sum_{k=0}^{\infty} (I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)||^k$

since i picked $p$ so that $||I-A_p|| < 1$, by a)

$||(I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)||^k$
Since this is esentially a geometric series which converges since $||I-A_p||<1$, it converges to $\frac{1}{1 - ||I-A_p||}$

Now $lim_{p \to \infty} A_p \to I,$ so $lim_{p \to \infty} \frac{1}{1 - ||I-A_p||} \to 1$
or $||\phi(A_p)|| \to 1$ as $p \to \infty$.

$||\phi(A_p) - 1 || \to 0$ as $p \to \infty$.

Thus $\phi(A)=A^{-1}$ is continuous at I.
 
  • #7
jakncoke said:
You can see the truth of c) by this

So for any sequence $A_n \to I$, we can always pick a $N \in \mathbb{N}$ such that $\forall p \geq N$, $||A_p - I||< 1$

In b), we proved that $A_p$ has an inverse. $A_p^{-1]}$ or the form
$A_p^{-1} = \sum_{k=0}^{\infty} (I-A_p)^k$.

Expanding it out a bit $I + (I - A_p) + (I -A_p)^2...$ since $||\sum_{k=0}^{\infty} (I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)||^k$

since i picked $p$ so that $||I-A_p|| < 1$, by a)

$||(I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)||^k$
Since this is esentially a geometric series which converges since $||I-A_p||<1$, it converges to $\frac{1}{1 - ||I-A_p||}$

Now $lim_{p \to \infty} A_p \to I,$ so $lim_{p \to \infty} \frac{1}{1 - ||I-A_p||} \to 1$
or $||\phi(A_p)|| \to 1$ as $p \to \infty$.

$||\phi(A_p) - 1 || \to 0$ as $p \to \infty$.

Thus $\phi(A)=A^{-1}$ is continuous at I.

But $\sum_{k=0}^{\infty }||I-A_p||^k\rightarrow\frac{1}{1-||I-A_p||}$ does not necessarily mean $||\sum_{k=0}^{\infty }(I-A_p)^k||\rightarrow\frac{1}{1-||I-A_p||}$

Do you mean $\frac{1}{1-||I-A_p||}\rightarrow1$ as $p\rightarrow\infty $ so $\sum_{k=0}^{\infty }||I-A_p||^k\rightarrow1$ so $||\sum_{k=0}^{\infty }(I-A_p)^k||\rightarrow 1$? But does $\sum_{k=0}^{\infty }||I-A_p||^k\rightarrow\frac{1}{1-||I-A_p||}$ mean $||\sum_{k=0}^{\infty }(I-A_p)^k||\rightarrow\frac{1}{1-||I-A_p||}$? I am not sure about this.
 
Last edited:
  • #8
jakncoke said:
You can see the truth of c) by this

So for any sequence $A_n \to I$, we can always pick a $N \in \mathbb{N}$ such that $\forall p \geq N$, $||A_p - I||< 1$

In b), we proved that $A_p$ has an inverse. $A_p^{-1]}$ or the form
$A_p^{-1} = \sum_{k=0}^{\infty} (I-A_p)^k$.

Expanding it out a bit $I + (I - A_p) + (I -A_p)^2...$ since $||\sum_{k=0}^{\infty} (I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)||^k$

since i picked $p$ so that $||I-A_p|| < 1$, by a)

$||(I-A_p)^k|| \leq \sum_{k=0}^{\infty} ||(I-A_p)||^k$
Since this is esentially a geometric series which converges since $||I-A_p||<1$, it converges to $\frac{1}{1 - ||I-A_p||}$

Now $lim_{p \to \infty} A_p \to I,$ so $lim_{p \to \infty} \frac{1}{1 - ||I-A_p||} \to 1$
or $||\phi(A_p)|| \to 1$ as $p \to \infty$.

$||\phi(A_p) - 1 || \to 0$ as $p \to \infty$.

Thus $\phi(A)=A^{-1}$ is continuous at I.

And what about d)? Thanks.
 
  • #9
For d), use the very detailed hints that are provided. If $C$ is close to $A$, then $K$ is small and therefore so is $H$. Thus $I-H$ is close to $I$ and by c) so is $(I-H)^{-1}$, from which $C^{-1}$ is close to $A^{-1}$.

But the hint that says $C^{-1}=A^{-1}(I-H)^{-1}$ seems to be wrong. I think it should be $C^{-1}=(I-H)^{-1}A^{-1}.$
 

FAQ: Proofs about invertible linear functions

What is an invertible linear function?

An invertible linear function is a mathematical function that can be represented by a straight line and has a unique input and output for each value. It is also known as a one-to-one function and has a inverse function that can "undo" its effects.

How can I determine if a linear function is invertible?

A linear function is invertible if it has a non-zero slope. This means that the function does not pass through the origin and has a unique output for each input. In other words, the function is not "flattening out" or becoming horizontal.

What is the inverse function of an invertible linear function?

The inverse function of an invertible linear function is the function that "undoes" its effects. It is found by swapping the input and output variables of the original function and solving for the output. It is represented by f-1(x).

How do I prove that a linear function is invertible?

To prove that a linear function is invertible, you can show that it has a non-zero slope, as mentioned in the answer to question 2. Additionally, you can show that the function is one-to-one by using the horizontal line test. This means that a horizontal line can only intersect the function at one point.

Can a linear function have an inverse if it is not invertible?

No, a linear function must be invertible in order to have an inverse. If a linear function has a slope of 0, it will be a horizontal line and will not pass the horizontal line test. This means that it is not a one-to-one function and does not have an inverse.

Similar threads

Replies
19
Views
2K
Replies
10
Views
1K
Replies
5
Views
1K
Replies
4
Views
2K
Replies
1
Views
1K
Replies
2
Views
1K
Replies
4
Views
681
Back
Top