« first day (2507 days earlier)      last day (2531 days later) » 

12:09 AM
you want to integrate $1/(1+z^2)^m$ still?
close the boundary in the upper half plane you need the residuum at $i$, your function is $\frac1{(z-i)^m}\cdot h(z)$ where $h(z)=\frac1{(z+i)^m}$ which is analytic at $i$
if put in a taylor expansion of $\frac1{(z+i)^m}$ at $i$ you find that the term of order $m-1$ will give you the residual
this term is $\frac1{(m-1)!}\frac{d^{m-1}}{dz^{m-1}}\frac1{(z+i)^m}$ evaluated at $i$
now the $m-1$st derivative of $(z+i)^{-m}$ is $(-m)(-m-1)\cdot...\cdot(-2m-1)(z+i)^{2m-1}$
 
To clarify what I was getting at with the citation of the Leibniz formula earlier
You want to evaluate $\displaystyle \frac{1}{(m-1)}\frac{d^{m-1}}{dz^{m-1}}(z+i)^{-m}(z-i)^{-m}$ at $z=\pm i$, I presume?
 
without the $(z-i)^{-m}$ term
 
Eh, you still need that if you want to get the residue.
 
not if you want the residue at $i$
 
12:17 AM
i sketched it briefly just now how you can see how to get the right formula
 
Yeah, I forgot that when you do the residue formula you have that multiplication
 
well im going to sleep, i miscalulated the derivative up there, but if you do it correctly it should be simple to get the residue
 
12:33 AM
oh I went afk lets see what people said :)
yeah I got the form $(-m)(-m-1)...(-2m-1)(z+i)^{2m-1}$
but i didnt see how it simplified
and what did s.harp mean "not if you want the residue at $i$"
maybe something cancels with the $(m-1)!$ in the denominator
 
that was to me.
 
I thought you just have the power one less, so you use $(z-u)^{-m-1}$
 
I was forgetting that the residue formula involves a multiplication by $(z-i)^{m}$.
 
oh, oops
right
okay so I have $\frac 1 {(m-1)!} (-m)(-m-1)\cdots(-2m-1)(2i)^{2m-1}$, for example
the term with $i$ with simplify depending on $m \bmod 4$
 
True, but you probably should leave those alone for now.
 
12:41 AM
so the question is how to simplify the rest in terms of $m$
okay sure
 
For the stuff in front, note that there are a total of $m-1$ factors that've been brought down
So you can pull out that many minus signs from it
 
I dont follow
 
You've got $(-m)(-m-1)\cdots(-2m-1)$ in front.
How many minus signs are in there?
 
$m-1$
 
12:44 AM
okay
 
So that's $-(-1)^{m}m(m+1)\cdots(2m+1)$.
Can you write that using factorials?
 
$(-1)^{m+1}\frac{(2m+1)!}{(m-1)!}$
 
Right.
 
:)
and then we have another $\frac 1 {(m-1)!}$
but this is already much better
 
Right. A pity that's not a bit different; it's nearly a binomial coefficient. Oh well.
 
12:47 AM
the original question was:
$f_X(x) = \frac C {(1+x^2)^m}$ is a probability density function for some values of $C$. what are they?
 
the professor said
he realizes that the homework question is harder than anything he's taught in class
so just do what we can
so if I just find a function of $m$ and say $C = \frac 1 m$ it will be good enough to show my effort
 
For reference, Mathematica returns $$\int_{-\infty}^\infty \frac{dz}{(1+z^2)^m}=\sqrt{\pi}\frac{\Gamma(m-1/2)}{\Gamma(m)}$$
Where $\Gamma(m)$ is the Gamma function
 
ooh that's pretty
 
It may well be equivalent to what you've got, in fact.
First, note that $\Gamma(m)=(m-1)!$.
 
12:50 AM
sure
 
So that already looks productive :P
 
lol THAnks
 
Also, we've got $\Gamma(1/2)=\sqrt{\pi}$.
So what's on top is equivalent to $\Gamma(1/2)\Gamma(m-1/2)$.
And I'm almost positive I've seen that before.
Actually, more simple
$\Gamma(m+1)=m\Gamma(m)$, if I'm remembering Gamma right
 
Yup
that's how we analytically continue it to every nonnegative integer from the integral definition
 
So $\Gamma(m-1/2)=(m-3/2)(m-5/2)\cdots(1/2)\Gamma(1/2).$
 
12:52 AM
like a factorial
 
Right.
So that gives $\pi(m-3/2)(m-5/2)\cdots(1/2)$.
Now, you'll get a factor of $2\pi$ from the residue theorem.
So that's plausible enough.
 
that makes sense
this is cool stuff
 
If it works, at any rate.
 
I mean, it should, unless we're making wrong assumptions about differentiability or something
 
right.
 
12:55 AM
but it's "obvious" that $f_X$ is analytic everywhere except for $\pm i$
 
yeah.
 
anyway, fun stuff to think about
 
Well, so long as it's integrable. That requires $m>1/2$.
 
even if us probability students are not equipped to do the whole thing
 
If $m=0$, then that obviously fails.
 
12:56 AM
oh
 
But then you'd have $f_X=1$ for all $x$ :)
 
which doesnt work!
 
Right-o.
The one thing I'll say for the approach you're doing: It only works for integer $m$.
 
yeah I realized that, but I dont know enough to do other $m$
 
But the result Mathematica gives is for all $m>1/2$.
Yeah.
 
12:57 AM
contour integration isnt even a prereq for this class, hence the professor's allowance
 
I think I know how to do it, or at least how to start.
First, by symmetry we can restrict the integration to x>0 at the cost of an overall factor of two.
 
because it's even
 
Hmmm
Right.
 
could someone please helpe me with this integral since i have no clue on how to start $$\int_{-\infty}^{\infty} e^{-(x-y)^2/4t - y^2} dy$$
 
Something something key-hole contour.
I know I've done integrals like that in the past, but eh.
 
12:59 AM
how do you know that's even integrable semiclassic
oh
I was thinking analyticity
you dont need analytic
 
Well, the fact that Mathematica gives a result gives me some confidence :)
 
oh I thought you were answering Alek
 
Ah.
@AlekMurt The golden rule of Gaussian integrals: Thou shalt complete the square in the exponent.
 
you sure that's a single integral, Alek?
 
well, x is any real, t is any real positive, it is just an integral that gives solution to the heat equation
 
1:01 AM
oh okay
 
i already saw the result on mathematica
 
Which is to say, you should seek to write your exponent in the form $-a(x-b)^2+c$ for appropriate constants $a,b,c$.
 
but i dont know how to solve it
 
Once you've gotten it in that form, your integral is $e^c\int_{-\infty}^\infty e^{-a(x-b)^2}\,dx$.
The $b$ can then be gotten rid of by an appropriate shift of the integration, and what's left is just a standard Gaussian integral.
 
thats the thing, i have tried that, but i have not reached that perfect square
 
1:03 AM
Okay.
Let me change that slightly for convenience and swap $c\to -c$.
Oh. I should also be using $y$ not $x$ since that's the integration variable.
In that case, you want $a(y-b)^2+c = (x-y)^2/(4t)+y^2$. Agreed?
 
ok
yep, thats what i want
 
If I expand the RHS, the coefficient we get for $y^2$ is $1+1/(4t)$.
Matching to the LHS, we have $a=1+1/(4t)$.
Similarly, if we match the $y$ terms on both sides we get $-2ab=-2/(4t)\implies b=1/(4at)=\frac{1}{1+4t}$
And finally for the constant term we have $ab^2+c = x^2/(4t)\implies c = x^2/(4t)-ab^2$.
The algebra for that is a bit tedious, eh.
 
oh men
 
hardly insurmountable, though.
 
i really didnt know about that trick, i mean, the way how u did it all
 
1:09 AM
btw, there's a way to get Mathematica to do those identifications for you. lemme scrounge it up
 
i always worked on another fashion but i think yours is way better, i really appreciate it man
 
Right. If you use the SolveAlways command, you can force it to solve such equalities under the assumption that it's true for all $y$.
So SolveAlways[a (y - b)^2 + c == (x - y)^2/(4 t) + y^2, y]
gives {{c -> b^2 (1 + 4 t), x -> b + 4 b t, a -> (1 + 4 t)/(4 t)}}
Unfortunately, I don't know how to make it choose which variables it solves in terms of.
But that gives $a=\frac{1+4t}{4t}$, $x=b+4bt\implies b=\frac{x}{1+4t}$, $c=b^2(1+4t) = \frac{x^2}{1+4t}$.
typo, fixing
Right. This also exposes a typo I made earlier.
should have been $-2ab=-2x/(4t)\implies b = x/(4at)=x/(1+4t)$.
 
oh i see, thank you man, i really appreciate it
 
oh hey, I figured out how to make it choose. yay for actually reading documentation
Solve[! Eliminate[! a (y - b)^2 + c == (x - y)^2/(4 t) + y^2, y], {a, b, c}]
gives
{{a -> (1 + 4 t)/(4 t), b -> x/(1 + 4 t), c -> x^2/(1 + 4 t)}}
(SolveAlways for y is apparently equivalent to doing the above code without {a,b,c} specific. I just added those in.)
(...that could have made some of my older calculations a lot simpler. Why did I never use that?!?)
 
1:31 AM
> Eliminate[!
Exterminate!
 
1:49 AM
possum
 
another question that I don't get.
that I probably should
 
Distance up, distance down.
 
that's what I said
so it's equal to the max height + (max height - height of building)
right?
 
yep.
Do you know how to find the max height?
 
We can complete the square
Which, I've learned today!
 
2:02 AM
yep, that works.
 
is there a better way?
 
you can also compute the velocity (the first derivative w/r/t time)
What's the velocity when it's reached its max height?
Hint: If it had a positive velocity, it'd still be moving up :)
 
0
:)
 
Right.
 
oh
 
2:04 AM
So you can do it that way too.
 
so find the first derivative
make it equal to zero
 
Yep. Both give the same result.
 
Noice.
I love doing math with you Semi.
 
That's not a bad exercise, in fact. Find the max of $y=ax^2+bx+c$ by both completing the square and by setting the derivative to zero.
You'll get the same value of $x$ in either case.
okay, I'm out for a bit
 
Okay, cya Semi.
Same time same place tomorrow.
Hey dami
hows it going.
 
2:12 AM
Hey there!
 
Doing homework?
 
Nope, done with classes, just kinda doing random stuff before bootcamp
 
bootcamp?
joining the army?
 
The analysis bootcamp
 
oh wow
I'm having trouble with integrals
I have a calculus text by wylie, and it does a really bad job of explaining
though, of course I don't need it for my test on monday, so I'm not worrying too much about it
but it explained it as $\frac{1}{n}[\frac{1^2}{n^2}+\frac{2^2}{n^2}+\frac{3^2}{n^2}....\frac{n-1}{n^2}]‌​$> area > $\frac{1}{n}[\frac{1^2}{n^2} ... \frac{n^2}{n^2}]$
and then the excersises were like "use this method to determine the area under the curve between x =0 and x = 2 for x^2
 
2:26 AM
@Dodsy I consider myself proficient at integral taking. What specific integral technique are you having trouble with?
 
Read above.
 
wat
makes confused face
you want to take the integral $\int_0^2 x^2 dx$
right?
 
right, sure.
 
ok
do you know the second fundamental theorem of calculus?
I'll assume you don't since that was a long pause. :-)
The second fundamental theorem of calculus states that if you select any single antiderivative of $f(x)$ and call it $F(x)$, then $\int_a^b f(x) dx = F(b) - F(a)$
@Dodsy so... what is the expression by which you describe the set of antiderivatives of $x^2$. Or for that matter, can you find any antiderivative for $x^2$?
@Dodsy im waiting...?
 
2:43 AM
Why is functional analysis so hard?
 
it shouldn't be
If you think it is hard then it shall be hard. If you believe it will be easy, then to you it shall become easy.
 
the proof of the unbounded spectral theorem is reallyyy long...
 
so?
does length = difficulty?
 
it's hard to understand all of it.
no, but length*difficulty = mental work needed
and mental work needed/max(time till test,1) = potential stress
and actual stress = potential stress iff potential stress < fuck this coefficient.
 
2:47 AM
otherwise stress = 0
 
now equate real and imaginary parts
 
1/mental work needed = potential stress
I embrace difficulty
it's trivial grunge work that annoys me
 
@GFauxPas are you saying my troubles are imaginary -.- ?
 
They're complex at least
 
My professor in FA said he really didn't try and understand the proof until after grad school when teaching it.
 
2:50 AM
hi-yo
 
Hi @semi
 
@MichaelBlane I liked to say in undergrad that a certain profs office hours were complex, in that they had both real and imaginary components.
 
@PVAL-inactive I have to prove it in front of the proffesor tomorrow, I can use the spectral theorem for bounded operators though
I have to prove existence of the spectral decomposition and my friend will prove uniqueness, and a lemma that we'll use also.
 
3:25 AM
Oh lord unbounded spectral theorem
I only know it for like, compact operators
 
4:03 AM
hi@Daminark
what is going on in analysis bootcamp?
 
It hasn't started yet, but it will be doing complex analysis, dynamics, geometry, and probability
 
I hope we can discuss what you learn from there :)
dynamics ?
like topics
 
Dynamical systems
 
nice
ok damin reboot your energy through bootcamp :)
btw how long is it?
 
Starts July 5th, goes until August 31st
 
4:12 AM
nice
 
4:23 AM
Okay stupid question: is $-x^2$ = $(-x)^2$ or $-(x^2)$?
In other which operation takes precedence, algebraically?
 
I would understand that as $-(x^2)$.
 
Lel
 
negation or exponentiation? Is exponentiation more "bound" to x or is negation more "bound" to the x?
It has implications with $\
 
Under the principle that $x^2=xx$, so $-x^2=-xx=(-1)(x)(x)$.
 
$\mathbb C$
yes, but there's also the princle that (-x)^2 = (-x)*(-x) under the substitution principle.
 
4:26 AM
Another reason to think like that is that $(-x)^2=x^2$, so writing it like that would serve no purpose.
 
^
 
except in the complex domain
 
Even in the complex domain that's true.
 
$(-1)^2 = 1$
So in $\mathbb{C}$ this still checks out
 
@Daminark Prove it. ;)
 
4:29 AM
You have $--1$, rotate one of the minuses and overlap it on the other to get $+1$
 
dude that's way better than the way Weil asked us to prove it
 
QED
 
such mathematicalness
 
:O
 
why are you bringing up quantum electrodynamics here
 
4:30 AM
@Semiclassical Because we're too scrubby to talk about QCD instead.
 
that's just silly
 
I'm actually lolling right now
 
In truth, I can barely talk about QED let alone QCD
 
I think you mean this $-$ $|$ and $-$ gives you $+$
 
I think I first heard QED to mean quantum electrodynamics and was confused hearing it at the end of a proof
 
4:31 AM
lol
 
@BAYMAX Ah yes, and -1 | 1, so that's where you get the vertical bar.
 
QCD should mean proof by contradiction :P
 
Mathematics is a collaborative effort.
@Semiclassical "Quod cerat demonstrandum"
 
Either that or proof by making a color picture.
 
(-i)^2 = -1
 
4:34 AM
hmm. would a proof by intimidation end with Q D:< ?
 
this is not equal to i^2
 
yes it is.
i^2=-1.
2
 
u sure?
 
woah, methinks you are a wizard
 
$x^2-(-x)^2 = (x-(-x))(x+(-x) = (2x)(0)$.
 
4:36 AM
@MARXOS $i$ and $-i$ are in fact algebraically indistinguishable quantities, which is to say that if any polynomial equation with coefficients in $\Bbb R$ has $i$ as a root, it necessarily also has $-i$ as a root.
 
For another approach.
 
This is true in general of conjugates.
Oh crap u rite
I'm leaving my comment as a testament to the fact that u was rite and I was rong
 
So $x^2=(-x)^2$.
 
hows semi?
 
@Fargle: excellent. thank you. I've never heard anyone say that they are (algebraically) indistinguishable quantities.
 
4:37 AM
not bad.
 
good =)
 
The above gives another demonstration of why $(x)^2$ and $(-x)^2$ are the same thing.
 
@Fargle huh?
Oh
 
@Daminark I had put "with coefficients in $\Bbb C$" at first, which has the easy counterexample $x - i = 0$
 
The counterexample to the previously stated thing was likely $z-i$
tfw sniped
 
4:38 AM
S N I P E D
tfw sniped about sniping
 
I was gonna say $(z-i)(z-2i)=0$, actually :P
 
Oh you got sniped in bringing up the snipe
Fuck this is gonna go on forever
 
oh god recursive snipe
 
lol
 
Let's just proceed by induction.
 
4:39 AM
LOL SEMI
 
an ouroboros of sniipe
 
@Semiclassical ????? what does benzene have to do with this /s
 
I don't know. I'll have to dream and find out.
Either that or stay up for so many consecutive days that I have a hallucinatory vision quest.
But sleep is probably faster.
 
@Semiclassical There are ways to do that without sleep deprivation.
Or so I hear.
 
@Fargle Or so you saw?
 
4:42 AM
@Daminark Shhhhhhhhshshshshshsh
 
Alright alright decibal count becomes negative
Also yo @EricSilva
 
ha ha yes @Fargle
 
Here's a tip: spend almost your whole day engaging on a problem and chance are part of it will start to appear in dreams
 
hey any idea of $\omega$ limit se ?
*set
If $C = C(t)$ is a trajectory of a time independent vector field, we
define the omega limit set of $C$, denoted by $ω(C)$, to consist of
all points $p$ such that there exists a sequence $t_{n} → ∞$ such that
$C(t_{n}) → p.$
 
I know someone who dreamed the proof of pigeonhole
 
4:47 AM
For me, dreams from intensive maths thinking the other day is known to create dreams that most mistaken them as me taking LSD
(Even though I never took any illicit drugs nor my kidney condition will survive it. I am dared to say if any illicit drug get into my body, I will drop dead in seconds)
I am pretty sure my 6 year of dream log is telling people that doing maths has the same effect on me as taking LSD by someone else
 
idk about you guys but truth is math in my mind feels like a pure drug haha
 
Maths is known to make my dreams super vivid. The mechanism is still under investigation though
Unfortunately maths that makes sense in my dreams are rare. The recent extraction of the left right middle associative tensor looking algebra is one of those rare things that might be consistent
Physics is also known to have similar effect. Perhaps jokingly speaking physics is a type of maths lol
:Troll face:
 
really? I don't believe that. Couple days ago I was ill so I couldn't do math like in 4 days. And during those days I dreamed about things, weird things and I was't doing math at all. However when I do math, I never dream about anything.
 
Guess you might have a opposite reaction to maths for your dreams compared to mine
 
Happens :) , when we do mathematics in our mind for a bit of time
usually I get ideas of an unsolved problem there previously some weeks ago
better we watch Inception :)
 
5:00 AM
That rare for me. Usually even my dreams failed to solve a problem, or when it did solve it, the memory is so unstable it drifted in seconds
@BAYMAX actually ever since I watched inception, my dreams start to become very complex with layerings false awakenings zig zags and other crazy stuff
 
ha ha :)
 
And sometimes after watching cloud atlas, my dreams became a self contained universe with memories and characters specific to that dream
Each dream has its own continuity ever since that point
 
my dreams tend to have quite a bit of continuity.
i just can't remember any of it when I wake :/
like watching really weird anime that vanishes into the air once i'm actually awake
 
Are they differentiable? @Semi
 
cultuvate that a bit more, and you might have a new problem solving channel. Our subconscious seemed to be pretty good at thinking outside the box
 
5:08 AM
maybe.
that really only works for stuff you know pretty well already, though
 
5:22 AM
@Semiclassical correct me if I am wrong but 18*64*6 < 13923, right?
 
yes.
by about a factor of 2.
 
to put that in perspective, the lesser value is how much data I need
including the allocation bookeeping
the larger one is what was actually allocated
in groups of 4 bytes
 
neat.
 
that is roughly 48 kilobytes of memory
for an 8*8*8 cube
i think i need to cut the memory down...
oh and dumping the memory went faster
 
yeah, that's a bit much
2
 
5:25 AM
i added messages so i could time it
it got slower as it went
the allocation was the bottleneck
i.e. the more that got allocated, the more it had to check where to shove stuff in
perhaps if i make the points NOT objects
it will go faster
 
5:47 AM
Rip, I have reached almost 100 days for the fanatic badge twice now, but yesterday I missed the log in because due to timezones, it wasn't quite a day yet when I logged in yesterday.
 
points object
i have made it to 25+ days in a row like 8 times now
havent got the 30 day badge yet
anyone know of a good book for learning pdes on ones own?
 
@Faust7 hmm?
@DemCodeLines wassup?
 
Partial differential equations
 
"points object"
you're not referring to me?
 
ah
 

« first day (2507 days earlier)      last day (2531 days later) »