« first day (3755 days earlier)      last day (1561 days later) » 
00:00 - 18:0018:00 - 23:00

00:46
Suppose $E(X)=\theta$. How come $Var((X-\theta)^2)=Var(X)$?
@schn That doesn't seem correct. Do you mean $\mathrm{E}\!\left((X-\theta)^2\right)=\mathrm{Var(X)}$?
That is pretty much the definition
This is on least squares estimators, when the sample is independent but not identically distributed.
@robjohn Maybe the second equality in the red box follows from using the estimator attained for $\theta$?
@robjohn Or it’s a typo somehow.
I think the square inside the first Var is not supposed to be there.
But I could be wrong
@robjohn But then, how is $Var(X-\theta)=Var(X)$? : )
Right, since Var(a)=0, for a a constant.
01:03
the variance measures how much the distribution deviates from the expected value, does that deviation change when we just translate the distribution by a given constant?
It doesn’t.
Yup
And that's precisely what $\operatorname{Var}(X-\theta)=\operatorname{Var}(X)$ says
heya @robjohn
@Thorgott But is $Var(X-\theta)$ related to $Var((X-\theta)^2)$, since the author specifically refers to the squared term?
no, I don't know what's going on there
01:07
Crazy.
@schn Be careful. Variance isn't linear.
@TedShifrin Good reminder.
But $\text{Var}(X) = E((X-\mu)^2)$ and $\text{Var}(X-a) = E([(X-a)-(\mu-a)]^2) = \text{Var}(X)$.
You actually don't need $a=\mu$. It's just that mean adjusts when you shift.
precisely, the deviation form the mean doesn't change when you translate the distribution, because you translate the mean by the same amount
which is exactly what that calculation expresses
I didn't even look at the stuff displayed. I just went according to what people said in here :P
01:15
So it is correct?
No, as robjohn said, the first Var should be E.
@TedShifrin Right.
@Ted can you give me an intuitive reason for why a curve fixed by an isometry is geodesic?
@Thor: Because locally geodesics joining two points are unique.
Think of reflection of a sphere across a plane through the origin.
Consider two (not too far) points on that great circle. If there were another geodesic joining those points, it and its reflection would give you two geodesics joining those points.
ah, an isometry necessarily preserves a geodesic joining two points and so if another curve joining those two points is preserved by an isometry, it must be that geodesic by uniqueness
01:21
Well, an isometry certainly can move a geodesic to a different one.
Just translate, rotate lines in the plane.
So be careful with what you said.
I think you're doing a converse or something.
OMG it's @Pedro !!!!
Doesn't even say hello. GRR.
That's better.
It's nice to see you're here.
Does anyone know how to show that if $p<q<r$ then for any $f$ we have $\| f\|_q \leqslant \max (\| f\|_p,\| f\|_r)$?
01:23
Oh, isn't there some sort of convexity?
This is all in $L^p(\mathbb R)$.
I assume you're on a finite measure space.
You have too many bars, don't you?
Yeah, I was asked about this by a friend.
Hm, let me see. My LaTeX is not rendering.
That should be better.
Yes, it's better.
I remember doing this studying for my analysis qualifying. Probably @Thor or @Alessandro knows this off the top of their heads.
Try this.
Yes, you've abandoned all your geometry, analysis, topology ... for formal algebra. You traitor!
Hahaha. I never pledged my allegiance to those. But I am teaching people about Fourier transforms and series and other people about L^p spaces. :)
01:28
OK, I won't curse you, then.
But my publications are algebra, mostly.
I still remember my first time ever visiting this chatroom was to discuss some multivariable analysis question you had. I'm not sure we ever settled it.
That feels like two lifetimes (and one pandemic) ago.
@Thor: Did you straighten it out (pun intended)?
Would anyone happen to know how to apply Regula Falsi to multiple dimensions?
I'm looking at this algorithm: stackoverflow.com/questions/36163846/…. Another answer in that Q&A describes the use of what's more or less Regula Falsi, however it isn't obvious to me how it could be used as a stand in for situations like this: stackoverflow.com/a/64719896/14073182 where there's nesting to multiple dimensions (although I know it can be done)
It's likely simple, I'm just rather awful at visualizing this sort of thing
I know it can be done, and I've searched high an low for an in-depth explanation, but all I can find are mentions of it... very frustrating
01:44
I think I got it. Pick $\alpha$ such that $\frac{\alpha}{p}+\frac{1-\alpha}{r}=\frac{1}{q}$. Then $p/\alpha q$ and $r/(1-\alpha)q$ are conjugate and Hölder gives $\int|f|^q=\int|f|^{\alpha q}|f|^{(1-\alpha)q}\le(\int|f|^p)^{\alpha q/p}(\int|f|^r)^{(1-\alpha)q/r}$. Taking the $q$-th root gives $\lVert f\rVert_q\le\lVert f\rVert_p^{\alpha}\lVert f\rVert_r^{1-\alpha}$. Lastly, observe $x^{\alpha}y^{1-\alpha}\le\max(x,y)$ for $0\le\alpha\le1$.
@TedShifrin What was the question?
@Thorgott Neat. ;)
math.stackexchange.com/questions/3906621 Can someone look at my thread and especially my comments if they have time?
@Pedro: I don't berember.
@Ted I think I understand what you meant by the reflections now. So if an isometry fixes two points (not too far), then, since it preserves lengths, it maps geodesics to geodesics and, since the endpoints are fixed, it fixes the geodesic between them by local uniqueness. But I'm not seeing how to get the converse.
02:18
If you had a different geodesic joining the points, its image under the isometry would give another. By uniqueness this cannot happen. Thus the fixed point set is totally geodesic.
02:31
Oh, I just realized the converse isn't true at all
but I agree the fixed point set is totally geodesic
ah ok, but if the fixed point is itself a curve, then, being totally geodesic, it is necessarily geodesic
02:50
Grr. Being told that my formula is wrong, but the person actually posted (what I think) is a wrong formula instead!
03:05
@TedShifrin howdy... just had dinner. How are things with you?
Sal
Sal
What's the general behaviour of $\sum_{n=1}^\infty x^n$ for $0 < x < 1$?
I'm pretty sure it converges to some value, but can that general value be expressed as anything other than an infinite sum?
@Sal other than $\frac{x}{1-x}$?
Sal
Sal
Perhaps.

I'm just not familiar with infinite sums. All I know is that a similar expression $\sum_{n=1}^\infty \dfrac{1}{2^n}$ is just $1$.
@Sal yeah, plug in $x=\frac12$
Sal
Sal
Is there a proof that I can look up that shows that $\sum_{n=1}^\infty = \dfrac{x}{1-x}$?
03:16
calculate the partial sums
what's $(1-x)\sum_{n=1}^kx^n$?
@Balarka thanks for linking that write-up btw, was a very instructive read
03:40
4
A: Congratulations: the big thread!

skullpatrolCongratulations Pedro Tamaroff for getting into the 100K club.

CC @robjohn
@skullpatrol yes?
congrats on getting into the club sir
@Sal Look up geometric series on Wikipedia. I bet there is a proof there.
@skullpatrol I've been in the 100K club for a while ;-)
Hullo
03:51
hi pal
I have a weird idea
We usually sum over the natural numbers or any other set of integers
But can we sum over an interval
The function to be summed should have the necessary conditions hold to make the sum converge
the sensible way of summing over an interval is called integration
Ya I know that
I just said it for a special reason
Like, to make silly students (who don't have much intuition for integration) think that it's a totally different
Lol
I have another idea
Suppose we a 2×2 matrix whose components are functions
How can we calculate it's determinant
I don't know what has happened to me why I am talking silly things
What is your opinion?
My opinion is that these silly writers write anything math is nothing without calculus
04:48
Happy Diwali Everyone
2
Happy diwali!
 
1 hour later…
06:04
@TedShifrin I mean I understood that at least one of principal curvature is zero. But how can I interpret it geometrically?
06:35
@love_sodam That means it is like a rolled up piece of paper, flat.
07:10
@love_sodam See p. 61 of my diff geo text. You'll find exercises about what possible developable surfaces you can have.
Your diff geo text?
@TedShifrin: Fortunately you are online!! I want to know more about your thoughts about math.stackexchange.com/q/3906461/272127
See my profile, @love_sodam.
@C.F.G I've said what I have to say. It's late here now.
@TedShifrin: OK, I'll wait until morning. sweet dream professor!
08:01
hello
I am back
08:22
in how many ways can we represent a number as a sum of squares?
@epic_math It depends on the number and its factorization
is there an exact formula for $n\in\Bbb{N}$ given the prime factorization of $n$?
I have a very interesting idea in my mind related to partitions.
$1105=4^2+33^2=9^2+32^2=12^2+31^2=23^2+24^2$
Note that $1105=5\cdot13\cdot17$ where each factor is $1\pmod4$
we know the formula for representing an integer as the sum of natural numbers, but what about representing as a sum of squares?
@Thorgott Glad to be of some use
08:37
$(2 + i) (3 + 2 i) (4 + i)=9+32i$
$(2 + i) (3 + 2 i) (4 - i)=23+24i$
$(2 + i) (3 - 2 i) (4 - i)=31-12i$
$(2 + i) (3 - 2 i) (4 + i)=33+4i$
can you tell me why you are multiplying complex numbers?
Each prime that is $1\pmod4$ is not a Gaussian Prime...
e.g. $5=(2+i)(2-i)$
so $5=2^2+1^2$
$13=(3+2i)(3-2i)$
$17=(4+i)(4-i)$
combining the various $+$s and $-$s gives different ways to write the product as a sum of squares
That's how I got the 4 ways to write 1105 as a sum of squares
I believe that the number of ways to write a number as a sum of squares is $2^{s-1}$ where $s$ is the sum of the exponents of prime factors that are $1\pmod4$
08:53
do you have a rigorous proof?
Pretty much using the factorization as Gaussian integers
If you are willing to write $25=5^2+0^2=3^2+4^2$ as two ways
0 is not to be included, btw, no need to write it
then the count will be different
yeah
that formula would be (I think) more complicated
have to do some research
You also need each prime factor that is $3\pmod4$ to appear an even number of times
08:56
this is why number theory is cool
also $2=(1+i)(1-i)$, so we include the exponents of $2$ in $s$
hey can you tell me what 'G.F.' means in OEIS?
Ah, no we don't count the exponents of $2$
GF? not sure
let me look...
hey there is a conjecture on the number of ways to represent a number as sums of squares
let $f(n)$ denote the number of such ways
@epic_math Looks like it means generating function
09:03
it is conjectured that $$f(n)\equiv cn^{\alpha}e^{\beta \sqrt[3]{n}}$$ where $$c=\frac{\zeta({3/2})^{2/3}}{\sqrt{3}(4\pi)^{7/6}}$$
and $\alpha=\frac{7}{6},\quad\beta=\frac{3}{2}\frac{\pi}{2}^{1/3}\zeta(3/2)^{2/3}$
don't know the symbol for ~ wrote it as $\equiv$
can you prove it?
and btw the generating function of $f(n)$ is $$\prod_{m=1}^{\infty}\frac{1}{1-n^{m^2}}$$
The idea in my mind was a 'generalized partition function' and here I am analyzing the partition as sums of squares
Let me ask it on MO I will link it here please see it
@epic_math That generating function will include $5^2+0^2$ as well as $3^2+4^2$
okay didn't notice
Let me think on that
so what I want is a formula not depending on the prime factorization
09:20
Oh... you are asking about any sum of squares... I was only thinking as the sum of two squares. Sorry
sums of two squares is very easy
That is $3=1^2+1^2+1^2$
I derived it myself
Yeah, when I saw the generating function, I realized that
your help was still good no worry
@robjohn in my question, I will ask about the distinct partitions, unlike what I mentioned here
09:37
would ask later don't I have much time today
maybe I will ask about non distinct partitions which are more important for me
I will do some research and post it in the analytic number theory room (owned by me) which is in MO.
 
1 hour later…
10:39
Hello
How can we prove this:
Where:
11:12
Analytic number theory proofs are sometimes hard
11:28
Can nobody solve it?
In mathematics, the Jacobi triple product is the mathematical identity: ∏ m = 1 ∞ ( 1 − x 2 m ) ( 1 + x 2 m − 1...
@epic_math ^
11:47
Good morning all
I have a small question where I am stuck on:
given this exp equation: $I = I_0 * e^(a*x)$ I want to solve for $a$:
I can either divide first by $I_0$ and take the $ln$ : $ln(I/I_0) = ax$
OR
take the $ln$ first and divide then: $ln(I)/ln(I_0)=ax$
The results differ then
I cannot find where I am wrong, they contradict
second way is wrong, you're breaking log rules
ln(I) = ln(I_0) + ax
subtract
Thank you very much.. I overlooked that one
you're welcome
12:13
@LeakyNun thanks!
If sigma(t,\theta) = (x(t)cos\theta,x(t)sin\theta,z(t)) is a surface patch (surface of revolution) where t->(x(t),0,z(t)) is a regular unit speed curve with x(t)>0,z'(t)>0, then in my computation, the Gaussian curvature is -1 if and only if x(t) = x''(t)
And I saw that the persudosphere (gamma(t) = (sin(t),0,cos(t)+ln(tan(t/2))) is also has Gaussian curvature -1
So I parametrized the persudosphere by arc length. the arc length was cot(\theta).
But x(t) = sin(cot^{-1}(t)), x(t)\neq x''(t)
What is the problem?
I will show my computation just a second
Now (x'z''-x''z')z'/x = -1 <-> x'z''z'-x''(z')^2 = -x and as x'x''+z'z'' = 0, (x')^2x''+(z')^2x'' = x so that x'' = x
e,f,g over comes from the second fundamental form
Oh it's pseudosphere and by the way in t\in (\pi/2,\pi) in that pseudosphere
12:47
@Merosity yes
It is from a density functional theory paper that goes a bit more mathematical on its foundations
ah right
that scary index thing lol
 
1 hour later…
13:51
are universal covers important for singular (co-)homology?
"Weak people take revenge, strong people forgive, and intelligent people ignore."-not albert Einstein
"and gods make you unable to ignore without gone rage"
Hullo
@Secret are you new to this room?
nope
13:55
Oh
I don't see you much in this room do I guessed that
I have been in some other communities
I used to be active before 2018
Okay
Lol
@epic_math it would be more helpful if you included all the conditions in the picture
@user2103480 the photo is not mine btw
It is from reddit
@epic_math yuck
@BalarkaSen at some point you gotta explain to me how to obtain any finitely presented group as the fundamental group of some space. The easy part is of course just the wedge product of each generator, but glueing it in the right way to obtain the relations didn't work out for me. I tried it by induction but failed so meh
It was an exercise of last semester's topology 1 course and that sounds a bit ambitious without substantial hints lol
14:16
@user2103480 no
@user2103480 yuck reddit is really yuck
@Thorgott in topological spaces, working with maps $I \times I \rightarrow X$ is way more chill than with $I \rightarrow X^I$. This has got something to do with exponentials not working well in that category, right? Like, in comparison to set theory, where I could curry the hell out of each function whenever I felt like it
@LeakyNun ok nice
Then bundles are probably also not so important
Different constructions of spaces are probably more important in the regard, to actually compute the stuff
14:32
In mathematics, in particular in algebraic topology, differential geometry and algebraic geometry, the Chern classes are characteristic classes associated with complex vector bundles. They have since found applications in physics, Calabi–Yau manifolds, string theory, Chern–Simons theory, knot theory, Gromov–Witten invariants, topological quantum field theory, the Chern theorem etc. Chern classes were introduced by Shiing-Shen Chern (1946). == Geometric approach == === Basic idea and motivation === Chern classes are characteristic classes. They are topological invariants associated with ...
complex line bundles = elements of H^2
That's an application of cohomology to line bundles tho and I think usernumbers was asking about the other way around
14:56
Yeh but no worries.
Lifting a homotopy between paths is the same proof as the one for lifting a path, just that we patch together squares, right?
Pretty much
Since you know the "left-side" of the square already, and you know lifts of paths are unique once you specify the start-point, the lift of the square is actually already determined by what you've already done
You're just checking continuity
And that can be done locally
Ok cool thanks
The general statement is that if p: E -> B is a covering space, and X x [0,1] -> B is a homotopy for which you have specified the lift X x {0} -> E, there is a unique lift of F to a homotopy on E with the specified starting-data
for X a compact metric space
Gotta use Lebesgue number somewhere here I believe
Actually I'm skeptical now, I think you can do this pretty generally
I only considered the case with X = [0,1] so it works in general there
15:30
a homotopy between paths is just a path in the path space
@LeakyNun that's the reason why I asked about exponentials in topological spaces; this is a useful analogy but isn't used in proofs of the more simple facts
You can curry when what you curry is locally compact
AKA Y l.c. => C(X x Y, Z) = C(X, Z^Y)
15:50
Ah cool, so can that be used to simplify proofs about homotopies and if not, what are the obstructions?
Right, $\mathbf{Top}$ isn't cartesian closed. The idea of currying is that we want to have bijections $\operatorname{Hom}(X\times Y,Z)\cong\operatorname{Hom}(X,Z^Y)$ (and the category theorist asks this to be natural in all variables). In more fancy language, this is requiring that taking products with $Y$ and exponentiating by $Y$ are adjoint functors and there are abstract nonsense reasons why this can't happen in $\mathbf{Top}$.
I believe algebraic topologists restrict themselves to the slightly adopted category of compactly generated weak Hausdorff spaces for this reason and slightly mo
compactly generated weak Hausdorff spaces are rather obtuse to define, but I believe they include every nice space you ever stumble across
Hello
Couldn't talk for a while
@user2103480 this seems like the right idea where did you get stuck
16:13
@BalarkaSen I would say I didn't even think it through enough to really get stuck. I know that by the construction that you showed me, the application of SvK, I can get A^nB^m=e for any two generators and n,m in Z. My next step would to just try an induction step to get three generators, but there I'm not so sure if I should glue three spaces at once in a way, or glue two spaces and then glue the resulting two that are left
16:25
@user2103480 No why would it be useful
I've never met a map that was easier to show is continuous as a map to a goddamn function-space
@Thorgott Nerd
@MikeMiller Instead of doing a slightly different glueing argument every time one might be able to do some shit like lifting paths in path spaces
It's exactly the same argument in both cases
I mean that's what currying says right
If you can curry then proving one means proving the other, you can immediately translate proofs
For translating proofs I need to check that I get a covering p:Y^I -> X^I and such
isn't that what you use the whole loop space suspension thing for
What is Z^Y?
C(Y, Z)?
16:30
So it's not obvious to me that the proofs are translated immediately
ye
it's the iNtERnAl hOm
why aren't you writing it as that?
I mean as C(-, -).
to make it more ambiguous
jk
C(-,-) is clumsy
cause it's supposed to be an eXpoNenTiaL oBJecT
@user2103480 Meh fine
I don't like this approach
16:33
@MikeMiller I quote Mike Miller: "I don't care"
Good man
@user2103480 Add a cell for every relator in the group
Use SvKT at each step, yeah
Why would you write C(-,-) and the exponential notation both in the same expression when they're supposed to mean the same thing?
Lmao C(X, Y^Z)
ok good point
Mike is officially silly
16:36
Its just confusing... or maybe there is some categorical justification for that which I don't understand.
Prove that C(-, X) is a sheaf in an appropriate setting
(Y^Z)^X is ugly
C(X, C(Y,Z)) is ugly
Use each symbol once per expression challenge
Looks like I joined chat at the wrong moment lol
16:38
yeah im leaving
well, I guess you could make the point it's supposed to be external hom, sometimes people write stuff like $\mathcal{C}(C,D)$ for the morphisms from $C$ to $D$ in a category $\mathcal{C}$
but in Top, when it works, you even have the adjunction internally, I believe
now I'm gone
Hello, does anyone know how to find spectral resolution of this operator?
I've never seen that Y lc => C(X x Y, Z) = C(X, C(Y, Z)). I know if the evaluation map C(Y, Z) x Y -> Z is continuous then, is continuous, then we have that homeomorphism.
that first equality is a homeo btw.
@user2103480 I quote Mike Miller: "Lets rob a bank".
Why does this look like hom-tensor adjunction that we have for modules (abelian categories even?)?
both are adjunctions with the internal hom
the internal hom in modules is adjoint to the tensor, the internal hom in nice enough top spaces is adjoint to the product
tensor and product are symmetric monoidal structures on those categories respectively
16:44
@feynhat one is external, one is internal hom
Also, category theory is one grand hidden project to just flex on the uninitiated
@MikeMiller no
also, I feel like you need hypothesis on $X$ and $Z$ as well, not just on $Y$, but I don't wanna think about it
@MikeMiller ^
@feynhat yikes. I type like a moron. No clue why I wrote 'is continuous' twice.
@Thorgott You don't dude as Feynhat says it relies on evaluation being continuous
ah, it seems we don't need any on $Z$
So long as locally compact means "for any x in an open U, there is a compact K in U with x in int(K)", it is a straightforward exercise to show the evaluation map is continuous
I put something like this on a midterm. For most people the hard part was showing that a compact Hausdorff space is locally compact in this sense.
16:50
nevermind, just checked, it does appear to in fact be true
To be clear C(X,Y) has the compact-open topology
Man just prove it it's not hard
Yeah. Its easy.
no i dont wanna do topology
I have to read Sard's theorem.
just lemme quote the nlab and enjoy the company of my own pretension
17:01
@feynhat Nobody can make you do that.
17:14
@Thorgott so you also failed no nlab november
was one of the first to fail
17:28
We've had the following theorem without proof in our lecture: Let $p \colon X \longrightarrow B$ be a fiber bundle, $f \colon (I^k, \partial I^k) \longrightarrow (B,b_0)$. Then there is an $\Tilde{f} \colon (I^k, \partial I^k \setminus I^{k-1} \times \{ 0 \}) \longrightarrow (X,x_0)$ with $p \circ \Tilde{f} = f$.
tilde without caps
widetilde
@Thorgott works for me, weird
oversquiggly
@EdwardEvans nice
17:36
huh
Is the proof worth looking at, or is it just "intuitive"? And how do I imagine this? I'd think of $p^{-1}(\{ x_0 \})$ as the "spine" of the fiber bundle, and the mapping's "slices" going continuously along that spine such the projection gives us the original mapping
The proof is exactly the same as the proof for covering spaces
You just lose uniqueness
Covering spaces are like twisty versions of $p: X \times D \to X$, where $D$ is discrete
Fiber bundles (what most kinds of fibrations are) are like twisty versions of $p: X \times Y \to X$, where $Y$ need not be discrete
This result is obvious for both of those
00:00 - 18:0018:00 - 23:00

« first day (3755 days earlier)      last day (1561 days later) »