« first day (754 days earlier)      last day (4266 days later) » 

11:00 PM
@JonasTeuwen Is this going to be an R-rated chat or G-rated chat?
 
@robjohn But there is no difference right... we just add 0!
 
@JonasTeuwen clothes on, clothes off; no difference? Perhaps there...
 
@robjohn Well, he does +1 and I do -1.
 
@JonasTeuwen Now I tried to prove last night that $S_N(f)(x) = \frac{1}{2}\int_0^x D_N(t) - 1 dt$ is uniformly bounded
 
@robjohn If you remove your clothes and then you add them again...
Looks easy...? But I cannot do it.
 
11:04 PM
@JonasTeuwen and now I split up the integral using the formula for the sine of the sum of two angles
@JonasTeuwen
 
Why?
Rescale it until it is trivial.
 
@JonasTeuwen And it turned out I have to find a uniform bound on $$\left| \int_0^{Nx} \frac{\sin \theta}{\theta} d \theta \right|$$
@JonasTeuwen Now I know that the limit as $N$ goes to infinity is $\pi/2$
 
Yeah.
 
@JonasTeuwen But does that tell me anything about uniform convergence?
 
@OldJohn I have an excellent beer.
Yes it does.
If you do it right.
 
11:06 PM
@JonasTeuwen what da ya mean?
I have a sequence of bounded functions
 
Dunno, my brains are beat. That's all I can say. But you have an expert here with a square...
 
@robjohn hey
@JonasTeuwen yes
 
@JonasTeuwen so you're saying that BenjaLim is wearing your clothes now?
 
@robjohn They are isomorphic.
 
@robjohn I have to obtain a uniform bound on the integral above
Now if I know that it converges to $\pi/2$ uniformly then I'm done
 
11:08 PM
That's just your standard singular integral that is not Lebesgue whatever.
 
@BenjaLim integration by parts...
just a sec...
 
DA FUQ???????????
 
Prove convergence using alternating series. I did it in an answer once.
Can do it a bit smarter.
 
@JonasTeuwen Analysis for me so far has been
obtain
estimate after estimate after estimate
@JonasTeuwen Did I tell you last night I evaluated the sine integral using Riemann - Lebesgue and the Dirichlet Kernel?
 
Limit of estimate is the value. if done right.
 
11:10 PM
@BenjaLim Here. Me and Rob.
The limit is very easy, but can do it in many ways.
 
ah shit
 
@JonasTeuwen When i solved the improper integrals like you did, I never explicity mentioned the second term like you have mentioned in your answer. Do you think it is necessary to do so? Do you think
\begin{equation}
\lim_{y \to \infty} \left (\int_{[y/\pi]\pi}^y f(x) \, \text{d}x \right ).
\end{equation}

may not converge to zero?
 
Yes, by oscillation.
Very small area but very large differences.
 
Ohh, I see, but then if it is oscillating, then it is not square integrable right?
 
It can still be.
Also, your limit is over the integer or not? Can't remember.
 
11:18 PM
@JonasTeuwen I can't apply Dini's theorem
 
Hmm, *methinks* the thing is like
if the improper integral exits, then the term goes to zero,
else the term may not go to zero, but then the actual improper integral itself does not converge.
 
The term does not go to $0$.
Are you sure about that?
How about $1_{\mathbf Q}$?
 
What is $1_{Q}$?
 
measures intelligence...
 
$$
\int_0^N\frac{\sin(x)}{x}\,\mathrm{d}x = \left.\frac{1-\cos(x)}{x}\right]_0^N+\int_0^N\frac{1-\cos(x)}{x^2}\,\mathrm{d}x
$$
 
11:22 PM
indicator function of the rationals
 
@robjohn How did you get $1 - \cos x$?
 
Ohh, I never heard it called by that name.
 
What... how?
 
@JonasTeuwen Will the improper integral for $1_{Q}$ exist? I don't think so.
 
Yes, it is $0$. (Lebesgue)
 
11:24 PM
@BenjaLim $u=\frac1x$ and $dv=d(1-\cos(x))$
 
Ah smart guy
 
robjohn the master of syntax.
 
You see I just did $dv = d(-\cos x)$
and then I had something that really messed up
 
Yes, but experience helps.
Where are the problems if you cut of the bloody integral?
Around $0$!
So the substitution would only switch around the problems if it runs up to $\infty$.
But if you cut it off they don't.
But perhaps robjohn had something more profound in mind.
@JayeshBadwaik If you want Riemann you can take the natural numbers.
 
@JonasTeuwen Nah, it just simplifies having to move the bottom limit of integration :-)
 
11:26 PM
@robjohn Sorry what do you mean?
Or any random non-dense subset of $[0,1]$ shifted will work as well if you are a bit careful.
 
@JonasTeuwen We could simply use $-\cos(x)$ if we moved the bottom limit of integration to $1$.
 
@robjohn So now I have that by looking at your integral above
 
@robjohn Yes, that is defining the problem away too?
 
The guy on the outside the uv term is
 
@JonasTeuwen Getting the point now.
 
11:27 PM
and $\frac{\sin(x)}{x}$ is good on $[0,1]$
 
$$\frac{ 1 - \cos Nx}{Nx} $$
 
@robjohn But you might want to estimate the $\sin$.
But whatever, I am exhausted :-).
 
@robjohn and we are left with $$-\int_0^{Nx} \frac{1 - \cos x}{x^2} dx$$
 
Let me finish my beers first.
 
@BenjaLim and that converges.
 
11:28 PM
yes.
But the problem is right from the start I know everything converges pointwise to $\pi/2$
I am looking for uniform convergence
 
@BenjaLim What exactly do you mean by uniform convergence?
 
Profound mental retardation so far reaching even small intellectual work requires extraordinary effort.
 
Well * a priori* I was looking at $$\int_0^{Nx} \frac{\sin\theta}{\theta} d\theta$$ yes?
I now call this $f_N(x)$ say
And I ask about whether $f_N(x) \rightarrow \pi/2$ uniformly
 
@BenjaLim I don't think it does. Look near $x=0$
 
Ok then stuff uniform convergence
I have been trying the whole of last night to get a uniform bound on this guy
 
11:31 PM
@BenjaLim what's one night? ;-)
 
Well yesterday afternoon as well
 
@JonasTeuwen Okay, last time. If a function is piecewise continuous and discountinuous at only a finite number of points, then my results will hold true right (about that integral going to zero if the improper integral converges?)?
 
Yes, I believe so. Perhaps even Riemann integrable is enough I think. That way it can only fail to be continuous outside of a null-set.
 
@JonasTeuwen Okay. Yup Riemann Integrable is enough I guess. Well, till now I had only Riemann Integrable functions to work with, and hence my confusion. Thanks.
 
Well... depends how you want to see the limit.
 
11:36 PM
??
 
Yes, if it is a $L^1$ function you can take $1_N$ too.
 
@JonasTeuwen Hmm.
 
Because that will be $0$ in $L^1$.
And then the limit is fine, as Riemann function it fails to have a limit pointwise (everywhere defined).
 
Yeah.
I have to adjust my brain to think outside compact spaces now.
It intuitively gravitates towards $R$ and $C$ and suitably well-behaved functions.
 
@JayeshBadwaik that's how one retains a shred of sanity :-)
 
11:43 PM
@robjohn hahaha :-)
 
user19161
11:55 PM
Oh dear, now I am falling in love with a third person on SE...
 
Will has lots of love to share 8-).
 
user19161
But Jonas is Love.
 
Jonas is Madness.
 
user19161
The first one, no hope. The second one, a mistake. The third one, just a fantasy perhaps.
 
user19161
@JonasTeuwen Same here!
 
11:57 PM
@WillHunting do you fall in love every time you change your nick?
 
user19161
@JayeshBadwaik No, I fall in love every day. :-)
 

« first day (754 days earlier)      last day (4266 days later) »