« first day (2502 days earlier)      last day (2814 days later) » 

14:03
Just out of curiosity, can anybody make sense of the picture in this link? link
it seems totally ridiculous
@s.harp lol
I don't know what a $C^*$ algebra is, but based on your question that's not the only difficulty in understanding the picture :P
looks like a kid drew it
@AlessandroCodenotti a $C^*$ algebra is a banach algebra with an involution $*$ so tjat $\|x^*x\|=\|x\|^2$. Every $C^*$ algebra is a sub algebra of $B(H)$ for some hilbert space $H$
I take it you also clicked on the picture, @s.harp ? There's a bit more of an explanation in the image description.
14:08
I know some of those words. No seriously, I don't have the prerequisites for that, but thanks for the explanation
@pjs36 I did, the description is also rather unwieldy and sort of difficult to see waht is meant. But I mainly asked out of fun because the picture looks so weird and almost stupid
Well, that's true -- it does have a "certain look" about it :)
That picture is what I'm gonna show my family when they ask what I do
They'll say, "Aw, sweetie, at least you're trying"
this is SERIOUS mathematics mom!!!
14:15
Here's another picture if anyone is easily amused by colors and graphs.
What's the cleverest way to integral $x^2/\sqrt{x^2+a^2}$
Wolframalpha
yeah well
Maybe not the cleverest, but surely the least painful :P
you could say that this is a 3d integral over $1/\sqrt{x^2+y^2+c^2+a^2}$ wit hsome constant
doing one integral gives you one of those inverse trigonometric guys
however integrating again is hard because of the $y^2$ so nevermind
14:18
Hey everyone
@s.harp Right...
Well, you could potentially do integration by parts with $u = x^2$ twice to reduce it down to $1/\sqrt{x^2+a^2}$ which integrates to $\log|x + \sqrt{x^2 + a^2}|$
but it's too much pain
I don't see why trig substitution would be so bad, or am I worse at calculus than ever before?
Like, $x = a\tan \theta$? You end up integrating $\sec^3 \theta$
Is there a clever way to do it? I think you have to go through a lot of IBP pain
I guess I am worse than ever, because I seem to end up integrating $\sec \theta$ (times a constant, maybe)
Sorry, you end up with $\tan^2 \theta \cdot \sec^2 \theta /\sec \theta$
$\tan^2 \theta \cdot \sec \theta$
14:26
Yeah, that's the one
That's terribad though
SBM
SBM
What is the the integral?
scroll up
SBM
SBM
@BalarkaSen Oh I guess hyperbolic things
And by parts
I can do it by parts, but I dunno if that's the easiest.
SBM
SBM
14:31
because well I guess $$\int \frac{1}{\sqrt{x^2 + a^2}} ~ \mathrm d x = \operatorname{arsinh} \left| \frac{x}{a} \right| + k $$
yeah something like that. arcsinh(x) = log(x + sqrt(1+x^2)) I think
which is what that integral is
SBM
SBM
I also am trying to solve a similar integral; which appears weird
but looks easy yet
I don't know the name of what I'm seeking, but maybe someone could help point me in the right direction. I do know that what I'm dealing with, at surface, presents as a programming problem but fundamentally is a math--possibly a combinatorics--problem. I'll as best I can lay it out programmatically.
Let's say I have an array containing 500 arrays, each containing unique combinations of precisely 6 values between 1 and 12.
var myArray = [ [1,2,3,4,5,6], [7,8,9,10,11,12], [2,3,4,5,6,7], ... ];
Now, let's say that I want to reduce this set of 500 arrays to a set of 20. But here's the catch: I
@MichaelP. I suspect what you're doing falls broadly under the heading of "combinatorial designs". I'm not an expert, but any time you have lots of sets and symmetry (they all contain this many, overall any element is included this many times, they all have this size overlap, etc), that's what comes to mind.
Oooh. I learned a new phrase today.
14:44
can someone give me a hint on this?
Unfortunately I found the terminology curve to be pretty steep for design theory, but it's worth a shot to look at this list and see if you can identify what "kind" of design you're after. Then, you can try and figure out what its parameters are, and there are huge databases of designs out there, that probably have what you need.
Of course, it may well turn out that it's just easier to figure it out yourself, than get sucked into terminology and literature etc. :)
@pjs36, that's LITERALLY the list I was viewing right before you messaged me. Ha.
@pjs36, maybe you're right. I'm not great at math, but this branch of mathematics at least provides a framework for me to investigate.
I was shooting blindly otherwise.
@MichaelP. I'm sure you're better at math than you think. :)
Yes, I know what you mean. Figuring out "what everyone else is calling the thing you're looking at" is often a nontrivial task -- it just happened to me, with something I've been looking at the last few days.
I think you probably have a block design, and there are a couple of formulas relating $v, b, r, k, \lambda$ (so many parameters!). I think for you $v = 12, b = 20, r = 2, k = 6, \lambda = ??$ (may not be possible)
ok, it seems I have to tune the definition of hyperoperations a bit in order to get it to behave like tetration without collapse. However, I am not sure if the larger ordinals can always successfully absorb smaller ordinals on the left...
14:57
I guess I meant $r = 10$, not $2$.
no hint? lol
15:44
@Secret what's the question we're exploring?
Please note the following is a work in process and I am on my way checking that it reproduce all standard results:
@Secret I'll help you check those
Associativity:
\begin{align}
\alpha[m<3](\beta[m<3]\gamma)=(\alpha[m<3]\beta)[m<3]\gamma
\end{align}

For $\alpha,n,\gamma < \omega$
\begin{align}
\alpha[0]\beta & =S(\alpha)=\alpha[1]1\\
\alpha[1]0 & = \alpha\\
\alpha[2]0 & = 0\\
\alpha[m>2]0 & =1\\
\alpha[m](\beta+1) & = \alpha[m-1]^{\beta}\alpha\\
\end{align}

For $\alpha,\beta,\gamma \geq \omega$
\begin{align}
\alpha[0]\beta & =S(\alpha)=\alpha[1]1\\
\alpha[1]0 & = \alpha\\
\alpha[2]0 & = 0\\
\alpha[m>2]0 & =1\\
\alpha[m](\beta+1) & = \alpha[m-1]^{\beta+1}\alpha = \alpha[m-1]^{\beta}(\alpha[m-1]\alpha)\\
The idea is to use the left continuity of infinite ordinals to mop up that extra term, while only keeping the $\omega+1$ height exponential tower alive (because right associativity should prevent the omega base from absorbing it)
I would write the first line as $$\forall m \in \{0,1,2\}: \alpha[m](\beta[m]\gamma)=(\alpha[m]\beta)[m]\gamma$$
I don't know about you, but I find the symbol $\beta^+$ for $S(\beta)$ quite convenient.
I found using $S$ as successor operator better for a personal preference, though it is pretty much rarely refer to after the rules are introduced
15:50
Are you sure that you want to write $$\alpha[m](\beta+1) = \alpha[m-1]^{\beta+1}\alpha = \alpha[m-1]^{\beta}(\alpha[m-1]\alpha)$$ instead of $$\alpha[m](\beta+1) = \alpha[m-1]^{\beta+1}\alpha = \alpha[m-1](\alpha[m-1]^{\beta}\alpha)$$?
Yes, as the top will reproduce cases like $\omega^{\omega+1},\omega^2+\omega,\omega +1$ note how there's always some lower ordinal surviving to the right of the larger ordinal due to the discontinuity of right operations
@Secret the top only works for $m \le 3$, as in all your three examples provided, because $m-1$ makes the hyperoperation associative.
and most importantly, the top case will ensure $\omega^{\epsilon_0}=\epsilon_0$ be avoided since the tower is evaluated from top down, this is needed to create a $\omega +1$ tower, (which is why I need to introduce the exponential concatenation earlier (whcih I have its rules updated, but as far)

This is precisely the point, the nonassociativity of $m=4$ prevent the tower from collapsing if I use the top defintion (because all operators will be evaluated from right to left
(Recall that all hyperoperators are right associative by definition)
No, if you use the top one, the collapse will begin even sooner.
hi @TedShifrin
15:58
Hi Karim
how is everything ?
Enjoying my European trek so far ... How're you?
good I have been busy I am working on understanding algebraic geometry, homological algebra, and triangulated categories. I am also solving Michael Atiyah. I would like by end of summer to have good grip to commutative algebra.
That puts you ahead of me, then ...
why its been so long since you thought about commutative algebra ?
16:03
Most of it, yeah.
oh I see
I bought Hartshrone. I find it very readable now since taking a course in commutative algebra.
I mean the course didn't cover everything that I need to understand the text, but some stuff I take for granted until I read the proofs for those things by myself later.
I never found it readable. Even taking the course from him.
How was his teaching style :D ?
Clear.
yeah his book sometime 1 page takes few hours. Before though when I used to read math text I used to read it like a novel, but right now I always have a pen and paper with me and make sure to cover every detail, so I understand things more clearly.
16:07
Hi @Ted
I noticed that if I don't do that math can sometime trick you that you understand something, but you don't really understand it.
I told you no novels years ago :)
Hi @Alessandro!
yeah :D
16:44
Solution to a differential equation plotted log linearly.
Involves the first zeta zero, and the GCD matrix.
I don't know if it means anything.
Here is the log linear plot of the solution to the same (differential equation) matrix without the zeta zero. Looks rather simple.
It seems odd that there's a different period for negative and positive inputs in the first graph.
@Fargle Yes I think so too
(*start*)
nn = 40;
T[n_, k_] :=
T[n, k] =
Which[n < 1 || k < 1, 0, n == 1 || k == 1, 1, k > n, T[k, n], n > k,
T[k, Mod[n, k, 1]],
True, -Sum[T[n, i], {i, n - 1}]];(*Michael Somos,Jul 18 2011*)
A = Chop[Table[
Table[N[T[n, k]/n^ZetaZero[1]], {k, 1, nn}], {n, 1, nn}]];
Total[MatrixExp[A t].(Range[nn]*0 + 1)];
Plot[Sign[Re[%]]*Log[Abs[Re[%]]], {t, -42, 42}]
(*end*)
$\alpha[m](\omega+1) \overset{\textrm{def}}{\not\equiv} \sup (\alpha[m-1]\beta|\beta < \omega+1)$

Because why not, I only introduce a pointwise discontinuity in this system.
what I did is basically force any supremum sequence that has exponential towers all of the same element except the topmost one, to become undefined

Now hopefully the layering functions $a_{\gamma}(\alpha[m-1]\alpha)$ actually form a well ordering and then comparison with the epsilon numbers should be possible by comparing the $a$s and the $\omega$
(if you see gaps in the above description, they will be filled in later...)
$m=0,1,2$ will survive because they can escape via associativity to reproduce the standard ordinal arithmetic
(in particular, the details about the $a$s will be filled in shortly...)
uh wait a sec... I need to revise how is $\omega(\omega+1)$ and $\omega^{\omega+1}$rigorously defined
perhaps they are actually defined via supremums since they are composed of limit ordinals
in that case, I will not be able to justify forcibly decoupling $a[4](\omega+1)$ from its supremums
17:49
Hello. I am having trouble with this problem: math.stackexchange.com/questions/1293310/…
Specifically, I am having trouble showing that $\mathscr{L}$ has the finite intersection property. Would someone help me, perhaps by responding to the comment I left on the linked question?
Also, why do we have $p \in [x,y]$ for all $x \in A$ and $y \in U$. How do you know that every $x$ is in $A$ and every $y$ is in $U$?
18:19
Question, can I show the absolute value of polynomial in [a,b] is bounded by the integral of polynomial in [a,b], using banach spaces?
Depends on what you mean banach spaces
Well, definitions from normed spaces. I was thinking of looking at $L_1[a,b]$
Or $L^1[a,b]$, depends on notations...
nvm the map $C([a,b])\to L^1([a,b])$ is continuous...
with lipschitz constant $|b-a|$
Hm, so in [0,1] constant 1?
yes
18:26
Interesting, will try to look at it
@MikeMiller I derived the Gysin sequence from Serre SS in things, but I am not sure if I understand how to describe the maps which splice the various exact sequences factoring through the $E^\infty$ terms are induced from the covering projection and integration over fiber
I mean the direct proof is $\int |f(x)| dx ≤\int \|f\|_\infty dx= |b-a|\,\|f\|_\infty$
Do you know how to do that?
@BalarkaSen I thought the relevant map was the transgression from like $E_{n,0} \to E_{0,n+1}$ or something.
Can it be bounded the other way around, with a constant multiplier of course?
18:27
And that would give you the element you cup with
serre is multiplicative so it's all determined by that
@Studentmath no, you can have functions that are very thin and reach $1$ at the peak. These have sup norm $1$ but the $L^1$ norm gets as small as you want
Polynomials?
Ah
you can approximate any continuous function as close as you like in sup norm by polynomials on compacta
@s.harp Okay, that gave me some to think about. Thanks!
@MikeMiller the differential should give rise to the $H^{p-n - 1}(B) \to H^p(B)$ map
18:32
With that I will depart for the rest of the week.. good weekend\week everyone!
but I don't see how to describe the $H^*(B) \to H^*(E)$ and $H^*(E) \to H^*(B)$ maps
They come from a homological algebra argument from SS
oh i see
There's some argument that you can write down a certain exact sequence using the first nonzero differential in the SS or something
I usually see this happen at the E^2 page or something...
I think this should be in McCleary
Does somebody here know things about primitive ideals?
I sort of want to advertise my question, which has very low views right now: math.stackexchange.com/questions/2317502/…
18:35
Reminds me, I've got a question of my own which I should put together.
@MikeMiller Hm, I'll look there
In general I remember people using this argument to prove that spin structures are obstructed by w_2 by writing down the corresponding exact sequence and then defining w_2 to be some image of the snake map
18:52
I forgot what spin structures are again. If the classifying map M --> BSO(n) of TM lifts to M --> BSpin(n)?
19:09
Yeah, don't worry about it.
There's gotta be a better place to see this argument carried out.
So, here's a sketch of the question I've got.
Consider the multi-valued Lambert-W function with $k$th branch $W_k(x)$. For real $x>-1/e$, the principal branch $W_0(x)$ is the root $w$ of $x=we^w$ for $w>-1.$
At $x=-1/e$, there's a square-root branch point; if we start on the principal branch and approach the negative real axis, one gets $W_0(x)$ from above and $W_{-1}(x)$ from below. (So the principal branch is continuous from above on the branch cut $(-\infty,-1/e)$.)
123
123
Hello.
One point that's important is that $\text{Re}(W_0(x))=\text{Re}(W_{-1}(x))$ and $\text{Im}(W_0(x))=-\text{Im}(W_{-1}(x))$ for $x<-1/e$.
@MikeMiller Ah, thanks a lot! That seems like a useful piece of homological algebra.
Which means that $\frac{1}{2}(W_0+W_{-1})$ represents the common real part of $W_0(x),W_{-1}(x)$.
But there's one more wrinkle here: $W_0,W_{-1}$ are both real for $-1/e<x<0$, so $\frac{1}{2}(W_0+W_{-1})$ is real across the branch point at $x=-1/e$.
And it seem to be the case that this combination is smooth across the branch point.
So I'm trying to figure out why that's true and what this is known as.
19:31
hi, i need to show that $f(x,y,z) = 0$ defines a regular map around a fixed point $p=(x_0,y_0,z_0)$ s.t $\nabla f(p) \ne 0$. i managed to do that. now i need to prove that the tangent plane at $p$ is $\dfrac{\partial f(p)}{\partial x}(x- x_0) + \dfrac{\partial f(p)}{\partial y }(y - y_0) +\dfrac{\partial f(p)}{\partial z } (z - z_0) = 0$
That's equivalent to $(\nabla f)_p\cdot (\mathbf{r}-\mathbf{r}_0)=0$.
so i need to prove that $\nabla f(p)$ is orthogonal to the plane
so i have $f(g(x,y)) = 0$ around $p $ , correct ? @Semiclassical
Sounds right.
$g$ is the param. around $p$
now when i derivative this
i have $f'(g(x,y) g'(x,y) = 0$
but $g'$ is a matrix $3\times 2$
You're getting past what I can say anything useful about, I'm afraid.
19:35
what do you mean?
At the level of geometry, the formula makes sense. But I don't remember enough of how to do the algebraic details.
i showed that $\nabla f(p) g'(x,y) = 0 $ for some $x,y$ , is this enough ?
$x,y$ is fixed. i mean there is a point in $\Bbb R \ ^ 2$ which i know it happens.
And again, I'll have to shrug. I simply don't remember how these arguments go.
19:38
huh.
ok :P
someone saw the NBA finals yesterday ?
19:49
In a $\Bbb R^2$, let $G$ be an open set containing some point $x$. If $d(y,x)< d(z,x)$ for every $z \in \partial G$, does it mean $y \in int(G)$?
@BalarkaSen I think that might be what you need. I'm not comfortable enough with it to say though.
@Emolga yes, $\partial G$ is closed and the distance between a point and a closed set is strictly positive, let's say $m=d(x,\partial G)$, then $d(x,y)<m$ so there is a ball of radius between $d(x,y)$ and $m$ contained in $G$ and containing $y$
20:06
@LeakyNun
$\textbf{Tier 4: Tetration (More carefully)}$

Recall that there are two roadblocks to defining ordinal tetration:

\begin{align}
\omega^{\epsilon_{\alpha}} & =\epsilon_{\alpha}\tag{1}\\
\alpha^{\beta\gamma} & =(\alpha^{\beta})^{\gamma}\tag{2}
\end{align}

This means, if we define tetration in the usual (bottom up) way:

\begin{align}
{}^{0}\alpha & =1\\
{}^{\beta+1}\alpha & = \alpha^{{}^{\beta}\alpha}\\
\end{align}

we will quickly hit Roadblock $(1)$ for example:

$\sup(\{{}^n\omega \leq \omega^{{}^n\omega} = {}^{n+1}\omega|n\in\Bbb{N}\})=\{{}^{\omega}\omega \leq \omega^{{}^{\omega}\omega
Given $k = \mathbb{Q}_p$ and $k_b=k(\sqrt{b})$ for $b \in k^*$, does anyone know what the group of norms of elements $Nk_b^*$ is?
20:20
Greetings.
hi @TedShifrin
Hi Liad
good evening @Ted
(finally I can't get the timezone wrong)
Rehi Alessandro
hi @Ted
20:21
Yeah, you're good :)
Heya Eric
i have $\nabla f(x_0,y_0,z_0) \ne 0$ , so i have to prove that $\{f(x,y,z) = 0\}$ is a regular surface around $(x_0,y_0,z_0)$ , i did it. now i have to show that the tangent plane at $(x_0,y_0,z_0)$ is $\dfrac{\partial f (x_0,y_0,z_0)}{\partial x} (x-x_0) + \dfrac{\partial f (x_0,y_0,z_0)}{\partial y} (y-y_0)+ \dfrac{\partial f (x_0,y_0,z_0)}{\partial z } (z-z_0) = 0$
Hey everyone
You have two errors, Liad.
errors? where?
huh, $=0$
First, the equation needs $=0$. OK
20:25
and the second?
Hello!!! I have a question. We have that $p=3 mod 4$.

Why does it hold that $\sqrt{c} \pmod{p}= c^{\frac{p+1}{4}} \pmod{p}$ ?
@Ted I've been working through Clelland kind of casually since summer started, I think I like that book a lot, it's nice
Second, you should write $\dfrac{\partial f}{\partial x}( \dots)$. Do you see the difference, Liad?
huh, ok yes
I critiqued the exercises pretty carefully, Eric.
20:28
so how do i show it? i need to show $\nabla f(x_0,y_0,z_0)$ is orthogonal to that plane right?
Show it's orthogonal to any curve lying in the surface, Liad.
In this line from Milnor's book, shouldn't it be $g : U \to g[U] \subseteq M \subseteq \mathbb{R}^k$
yea i tried it.
im gonna go work on that and PDE now bye chat
20:29
we have $f(g(x,y) ) = 0$
so when i derivative it, we have :
$\nabla f(g(x,y) ) g'(x,y) = 0$
Bye, Eric.
am i right?
Curve, Liad. Curve.
Because since $g$ is a diffeomorphism it must be surjective, and hence that line implied that $g$ maps $U$ onto all of $M$
Hello @TedShifrin How are you? Did you see my question above?
We have that $p=3 mod 4$.

Why does it hold that $\sqrt{c} \pmod{p}= c^{\frac{p+1}{4}} \pmod{p}$ ?
20:30
No, @Perturbative.
but regular surfaces have a map $g:U\subset \Bbb R\ ^ 2 \to S$ doesn't it?
im not sure what do you mean by curve
I said to show the gradient is orthogonal to the tangent vector of any curve, Liad.
Why does $\sqrt c$ make sense, mod p, Evinda?
@Evinda do you agree that $c^{(p-1)/2} = 1$?
If there is a square root...
@TedShifrin Did Milnor mean that $g$ maps $U$ onto the neighbourhood $g[U]$ of $x$ in $M$, in the sense that if we let $N = g[U]$, then the map $g : U \to N$ is a diffeomorphism?,
Because unless I'm totally wrong, $g: U \to M$ is not a diffeomorphism
Right. But we do not yet know what diffeo means for manifolds, do we?
20:36
@TedShifrin if im working with a curve, we have $f(\gamma(t) ) = 0$ for all $t$ in some interval, then we have $\nabla f(\gamma(t) ) \gamma'(t) = 0$ for all $t$ , and now we can set $\gamma(t_0) = (x_0,y_0,z_0) $ and have that $\nabla f(\gamma(t_0) ) \gamma'(t_0) = 0$ , correct ?
@MoarCake559 If we consider $\sqrt{c}$ as an element of $\mathbb{F}_p^{\star}$ and taking into consideration that $a^{p-1}=1 \pmod{p}, \forall a: (a,p)=1$ we deduce that $c^{\frac{p-1}{2}}=1 \pmod{p}$, right?
@TedShifrin I found it in my textbook:
Yes, Liad :)
yes
this is one of the definitions for the Legendre symbol
wait, but in this way i have that $\nabla f(x_0,y_0,z_0)$ is only orthogonal to $\gamma'(t_0)$ dont i ?
20:38
and if you multiply c^{\frac{p-1}{2}} by c, then you get c^{\frac{p+1}{2}}
Since we know c has a square root, there is a well-defined c^{\frac{p+1}{2}}
True for every curve, Liad.
Yes, but how does this help? @MoarCake559
@TedShifrin I'm not sure I'm following what you're hinting at
c^{\frac{p+1}{2}} = c
what is the square root of both sides?
Ted!
how are you travels
20:42
I'm saying you're using language tgat doesn't yet make sense, probably, @Perturbative. It's been years since I've looked at Milnor. But what you say, uktimately, is correct.
Hey, Nate. Leaving Paris in 10 hours or so.
woah, awesome :)
You sound like you've had a few brewski's ;)
maybe some wine?
@TedShifrin Ohh, okay, I was worried that I had some error in my understanding
I do?
Typing on the iPad sucks. Tons of typos.
Oh, right.
@TedShifrin Thanks for the help!
20:44
Sure. :)
If I was in Europe, I'd be drunk most of the day. So perhaps I was projecting.
I've drunk but I'm not drunk.
@TedShifrin i think i got what you said, correct me please if im wrong:
the tangent plane is the union of all vectors that tangent to a curve $\gamma$ that is in $S$ and $p \in im(\gamma)$ , so i showed there that every such tangent vector is orthogonal to $\nabla f(p)$
Right.
great , thanks.
20:46
great, now I'm interested in the etymology of the word "drunk"
thanks, ted
German, Nate.
So where are you now then, friend?
bump: Given $k = \mathbb{Q}_p$ and $k_b=k(\sqrt{b})$ for $b \in k^*$, does anyone know what the group of norms of elements $Nk_b^*$ is?
Huh? Still in Paris. Going to Nice for 1 1/2 days.
ohhhh
"leaving paris in a 10 hours"
I read it wrong.
I have also been drinking.
20:48
Hey all :-)
Uh huh.
Hello, Kari!
How does one even multiply Kronecker deltas e.g. $\delta_{1i} \delta_{1j}$?
@TedShifrin are you going to bring me back trinkets?
@Kari: you're summing over what?
20:50
@Kari are you Norwegian? My girlfriend's mother shares your name and her parents were Norwegian.
Not likely, Nate. Your girlfriend woukd be jealous.
I've got that Inner product, @Ted
So I'm not entirely sure how to sum with the $i$ and the $j$
@TedShifrin haha, perhaps!
I would have to share, I suppose.
@Dodsy It's just the name of a character from a show, Hikari (shortened to Kari in the American version)
20:52
There are typos in that, @Kari. Things that shoukd be subscripts. Why are you multiplying deltas?
(It's a nice name though)
If $f_n$ converge uniformly to $f$ and $x_n \to x$, can I say $f_n(x_n) \to f(x)$?
Prove it, Emolga.
hey @Dair
Oops, I messed up. The last entries of the first terms should certainly be $x_i$ and $x_j$ respectively
20:53
@MoarCake559 Ah then we get that $c^{\frac{1}{2}}=c^{\frac{p+1}{4}}$, right?
@Dodsy hi
The deltas arise from $i$ varying in the $\partial_{x_i}$, @Ted
To everyone: what's that story that's almost a joke about a PhD candidate explaining something to some supervisors and the supervisors saying "in what field?" or something like that.
@Evinda Try writing out what I said on paper and then make a decision for yourself.
I'm trying to find it but am having a hard time.
20:54
If needed, take a break
they might mention $R^n$
I see. So you're doing $\sum_k \delta_{ki}\delta_{ki}$.
Yea, I really should've written it out like that!
Wait, that's a bit off. It should be $\delta_{ki} \delta_{kj}$, @Ted
@TedShifrin If this is true then I have a too-simple proof for the following:
if $f_n$ are analytic functions converging uniformly to analytic injective $f$ on a compact set $K$, then almost all $f_n$ are injective. Proof: Say $f(a_n)=f(b_n)$, Passing twice to a subsequence we can assume $a_n \to a$ and $b_n \to b$, then $f(a)=f(b)$ by what I wrote. But textbook says this follows from Rouche's theorem, so do you see a mistake in my proof?
No difference, Kari.
20:57
I did... we get that $c^{\frac{1}{2}}= \pm c^{\frac{p+1}{4}}$... @MoarCake559
@Emolga why is $\lim a_n \neq \lim b_n$?
Great, this shows where is my mistake.
So what is that sum, @Kari?
on the other hand i don't see a counterexample for continuous functions immediately

« first day (2502 days earlier)      last day (2814 days later) »