1. ## Arzelà-Ascoli theorem

Hi

Please, can some one give me lecture and exercises in Arzelà-Ascoli theorem ??

Miss_Lolitta

2. Originally Posted by miss_lolitta
Hi

Please, can some one give me lecture and exercises in Arzelà-Ascoli theorem ??

Miss_Lolitta

3. Why would a miss-"lolitta" be interested in such messy stuff?

Anyway. Here's one: Let $\displaystyle C^{0,a}$ be the set of Hölder continuous functions on $\displaystyle [0,1]$, with exponent $\displaystyle 0<a\leq 1$. Define the norm

$\displaystyle ||f||_a={\rm max}_{x\in [0,1]}|f(x)|+{\rm sup}_{x,y\in [0,1]}\frac{|f(x)-f(y)|}{|x-y|^a}$.

Show that the closed unit ball $\displaystyle ||f||_a\leq 1$ of $\displaystyle C^{0,a}$ is compact in the space of all continuous functions on $\displaystyle [0,1]$.

(Enjoy .
Taken from Royden's "Real Analysis" )

4. Really this is messy stuff!!

I wanted lecture in it because I have an assignment how to write a lecture with explaining by examples then represent this lecture in class..

and How I perform this a simply.. without complexity??

ThePerfectHacker & Rebesques....

Thanks so much

5. Well... If you want, you might say something like this. The Bolzano-Weirstrass theorem, sais that a bounded sequence converges. What can we say about a sequence of functions?

A problem is, we need to be precise about the mode of convergence - so consider the best possible, uniform convergence - the one implied by the sup-norm: $\displaystyle ||f||=\sup|f(x)|$

So, if we simply have a sequence of functions $\displaystyle (f_n)$, and it is bounded (i.e., the sup-norm is finite for all), do we have uniform convergence? That is, is there a function $\displaystyle f$ with$\displaystyle ||f_n-f||=\sup_{x}|f_n(x)-f(x)|\rightarrow 0$ ???

No. The domain could be unbounded, and so our functions may not approach a limit function (though they are bounded).

We fix this (and alike problems) by demanding the domain be compact, say the interval $\displaystyle [0,1]$

So, are we now to say uniform convergence is guaranteed? No, again. It could happen we only have pointwise convergence alone (not in the sup-norm, but only that for each $\displaystyle x$, $\displaystyle \lim_n f_n(x)=f(x)$.)

The additional hypothesis to resolve this, is (uniform) equicontinuity; That is, the $\displaystyle \delta$ in the $\displaystyle \epsilon-\delta$ definition of continuity, holds globally for all $\displaystyle x$ and $\displaystyle f_n$.

(Notice that that the Holder continuous functions in the last exercise do not satisfy this uniformly - but a relaxed form of the theorem can be used.)

So, every bounded and (uniformly) equicontinuous sequence of functions defined on $\displaystyle [0,1]$, necessarily has a (uniformly) convergent subsequence - this is what the theorem says.

Example. Take $\displaystyle \sin(nx)=:f_n(x), \ x\in [0,1]$. This is a bounded and equicontinuous sequence (show!). Note that for every $\displaystyle x, \ \lim_n f_n(x)$ does not exist (besides trivial cases).

But, by the Ascoli-Arzela theorem, there exists some subsequence $\displaystyle ( \sin(k_nx))$ and a function $\displaystyle f(x), \ x\in [0,1]$ with

$\displaystyle {\sup}_{x\in [0,1]} |{\rm sin}(k_n x)-f(x)|\rightarrow 0, \ n\rightarrow\infty$ .

Hope that helps.

6. That's perfect!!
Well done!!

Let K be a compact subset of Rp and let F be a collection of functions which continuous on K and have values in Rq. The following properties are equivalent:

1- The family F is bounded and uniformly equicontinuous on K.
2- Every sequence from F has a subsequence which is uniformly convergent on K.( this is another state for Arzelà-Ascoli theorem )

Thanks soooo00S00oooo much

7. Thanks soooo00S00oooo much
no need, love. i am just dead bored at work.

Let K be a compact subset of Rp and let F be a collection of functions which continuous on K and have values in Rq. The following properties are equivalent:

1- The family F is bounded and uniformly equicontinuous on K.
2- Every sequence from F has a subsequence which is uniformly convergent on K.( this is another state for Arzelà-Ascoli theorem)

1 implies 2: This is the proof of the Arzela-Ascoli theorem. It is kind of tedious, but at least there's a diagonal argument.

a) Consider all points of $\displaystyle K$ with rational coefficients. Call this (numerable) set $\displaystyle Q$, and denote its elements by $\displaystyle x_m, \ m=1,2\ldots$.

b) Let $\displaystyle (f_n)$ be any sequence of elements of $\displaystyle F$. We show it has a pointwise convergent subsequence on $\displaystyle Q$.

The sequence $\displaystyle f_n(x_1)$ is bounded; By the Bolzano-Weierstrass theorem, it has a convergent subsequence, say $\displaystyle f_{1,n}(x_1), \ n\in {\bf N}.$ Apply the same argument for $\displaystyle f_{1,n}(x_2)$, yielding a convergent subsequence $\displaystyle f_{2,n}(x_2), \ n\in {\bf N}.$ (note: the sequence $\displaystyle (f_{2,n})$ is thus convergent at $\displaystyle x_1$ and $\displaystyle x_2$). Proceed in this fashion, to obtain a subsequence $\displaystyle (f_{k,n})_{k,n\in {\bf N}}$ of the original sequence $\displaystyle (f_n)$. Then, the sequence $\displaystyle (f_{n,n})$ converges for all elements of $\displaystyle Q$.

c) We now show that $\displaystyle (f_{n,n})$ converges uniformly on $\displaystyle K$. That is, $\displaystyle f_{n,n}(x)$ is Cauchy for all $\displaystyle x$, in the uniform convergence norm.

Let $\displaystyle x\in K, \ m,n\in {\bf N}$ (these will be fixed later) and $\displaystyle \epsilon >0$. There exists some $\displaystyle x_i\in Q$ with

$\displaystyle |f_{n,n}(x)-f_{n,n}(x_i)| \ , \ |f_{m,m}(x)-f_{m,m}(x_i)|\leq \epsilon/3 \ (1)$

due to equicontinuity of the functions and the density of $\displaystyle Q$ in $\displaystyle K$. Also, we have

$\displaystyle |f_{n,n}(x_i)-f_{m,m}(x_i)|\leq \epsilon/3 \ (2)$

for sufficiently large $\displaystyle n,m$.

Therefore

$\displaystyle |f_{n,n}(x)-f_{m,m}(x)|\leq|f_{n,n}(x)-f_{n,n}(x_i)|+|f_{n,n}(x_i)-f_{m,m}(x_i)|$
$\displaystyle +|f_{m,m}(x_i)-f_{m,m}(x)|\leq \epsilon,$

by (1) and (2). So

$\displaystyle \sup_{x\in K}|f_{n,n}(x)-f_{m,m}(x)|\leq \epsilon,$

for all sufficiently large $\displaystyle n,m$; this means the subsequence is uniformly convergent on $\displaystyle K$.

2 implies 1: Let $\displaystyle C(K)$ be the metric space of continuous functions on $\displaystyle K$ endowed with the uniform convergence norm.

The given condition implies that $\displaystyle F$ is compact in $\displaystyle C(K)$, so it is bounded and closed in its topology.

We now show it is equicontinuous. Let $\displaystyle f$ belong to $\displaystyle F$. Then there is a sequence $\displaystyle (f_n)$ in $\displaystyle F$ converging uniformly to $\displaystyle f$.

Let $\displaystyle \epsilon>0, \ x,\ y \in K$ (to be fixed later). Consider $\displaystyle n$ so large that

$\displaystyle |f_n(x)-f(x)|, \ |f_n(y)-f(y)|\leq \epsilon/3 \ (3)$

and, from the uniform continuity of $\displaystyle f_m$, there exists
$\displaystyle \delta=\delta(\epsilon)$ such that

$\displaystyle |f_n(x)-f_n(y)|\leq \epsilon/3 \ (4)$

for $\displaystyle |x-y|\leq \delta$. We thus obtain

$\displaystyle |f(x)-f(y)|\leq |f_n(x)-f(x)|+|f_n(x)-f_n(y)|$
$\displaystyle +|f_n(y)-f(y)|\leq \epsilon$

due to (3) and (4). So $\displaystyle F$ is uniformly equicontinuous.

8. Really I don't know how can I thank ..

I'm grateful for you

I repeat...

Thanks soooo00S00oooo much

9. Originally Posted by Rebesques
no need, love. i am just dead bored at work.
I'm so sorry

10. I'm so sorry

Not as sorry as I am.

Do come back with more theorems like this.

11. ok

good luck

12. That's nothing like the other one.

13. I know that

but you understand it on the fly

try again to take bonus!!

14. try again to take bonus!!
Sorry... I can't be bribed.

(...unless it's really much.)

15. oh,you can't be bribed..

but bonus was candy..Do you like this bribe??

Rebesques,,I would like to tell you that presentation I presented it to my professor who will say the result is it good or no on next Monday ,,

but I'm sure it's excellent!!

just because Rebesques helped me without any bribe..

regard..

lolitta