« first day (3832 days earlier)      last day (1487 days later) » 

00:00
@TedShifrin: I have my gravatars for Valentines' Day, St Patrick's Day, and Easter lined up.
@TedShifrin I see, that clarifies it. So $Ax=b$ would be the BVP and then by introducing $G$, we can cleverly write $x=A^{-1}b$, i.e. we have an explicit expression of $u$ which only demands finding $G$.
I'm skeered @robjohn
peer pressure works!
amWhy cannot make me cower!
Ted the unshakable.
00:01
flaps wings
@TedShifrin Not cowering. I'm having fun making the gravatars to fit me.
Oh, I never intimated you had cowered. You're having a blast, @robjohn.
So far, I have Halloween, Thanksgiving, Christmas, Valentines Day, St Patricks Day, and Easter.
ROFL
You'll need another outfit next year!
@TedShifrin I hope to simply remove the mask, but we'll see.
00:06
I never did mask my dreidl.
@TedShifrin had a snack and came back. This seems interesting. So what is the sort of "minimal" required structure? that all ideals be principal?
Definitely not, methinks, @BigSocks.
Might be a chore. I have code to generate something close to your dreidel, but it doesn't hit it exactly.
I did three parametric plot 3d’s in Mathrmatica.
28 years ago or so.
I posted this on chat a while ago.
00:11
Looks great to me! What's wrong?
@TedShifrin hmm perhaps it also requires some other property $\Bbb Z / n \Bbb Z$
Quotient of PID isn't gonna work, Big.
@TedShifrin Nothing is wrong, it's just not exactly your shape, but maybe I'll mask it
Really? Those aren't cones over hyperboloid of one sheet?
Ted's is just spinning
also lol I just understood it was meant to be a dreidel
00:13
Nah. It was a solution to a math question.
@TedShifrin you mean in general
Yeah, I think.
@TedShifrin oh lol ok nice
See if you can prove that case.
$\Leftarrow$ is obvious so I gotta do $\Rightarrow$
00:15
Leftarrow is obvious in any ring!
yeah, moreover
@TedShifrin Let me know if I still misunderstand any of the points you made earlier. Thanks for the insight.
so $a = rb$ and $b = sa$ for $r,s \in \Bbb Z/ n \Bbb Z$. subbing one in the other you get $b = srb$, so $sr = 1$ and I think that does it
@schn You must mean the dude, @TedShifrin
@BigSocks Of course, sorry :)
Nope, Big. $b$ is a zero-divisor, right?
00:22
huh. hmmm
wait why
I thought it was just for arbitrary ideals $(a) = (b)$
I don't remember $b$ being a zero divisor
In $Z/n$ you’re either a unit or a zero-divisor.
hmm ok I buy this
Don't buy it, prove it.
You know how.
dang, I really should...
Then fix your non-proof :)
00:48
slightly ironic that st patrick's day is celebrated by large intake of liquids.
Your heritage!
hmm ok well $v$ is a unit whenever $\exists u [ vu = 1]$ and $v$ is a zero-divisor whenever $v \neq 0$ and $\exists u [vu = 0 \wedge u \neq 0]$.

Assume $v$ not a unit and it's not just $0$. Then $\neg \exists u [vu = 1]$. So $\{ vm \vert m \in \Bbb Z / n \Bbb Z \} \subsetneq \Bbb Z / n \Bbb Z $. So left multiplication is not onto. I guess you can pick elements $vm, vl$ such that $vm = vl$ in $\Bbb Z / n \Bbb Z$, so $vm - vl = 0$, and $v(m - l) = 0$, and neither of these are $0$ so $u = m - l$ and $v$ a zero-divisor.
@TedShifrin when we were growing up it was mostly a sober affair where you went to mass and wore a sprig of shamrock on your lapel. mighty imbibing was heavily frowned upon.
the first time i had a drink on st patricks day was in ireland's 32 in sf.
01:24
@Ted @Fargle There are really neat examples in $C(\mathbb{R},\mathbb{R})$ (where multiplication is pointwise multiplication). Units in this ring are just nowhere-zero functions, i.e. those of constant sign, so two associated functions have the same sign everywhere. On the other hand, for $g$ to divide $f$ is just to ask that $g$ is non-zero whenever $f$ is non-zero.
So take two functions having the same sign on $(-\infty,0)$, have them be zero on $[0,1]$, then let them have different signs on $(1,\infty)$. They divide each other, hence generate the same ideal, but aren't associate.
@Thorgott Are you saying x^2 divides x?
oh, you're right
the problem of dividing is more complicated due to continuity issues
but it's clear that examples can be constructed this way regardless
You want the limit of f/g at zeroes of g to exist
should read before replying.
just take $f=g$ on $(-\infty,0)$ and $f=-g$ on $(1,\infty)$
01:33
Yeah, I agree with that
01:45
Can someone show how the right hand side in the end is $O(e^2)$
this is steffensen's method
@Thorgott That's cute, I like that.
I made a typo in the end, the RHS of the last line is $e(1-\frac{O(e)}{O(e)}$)
02:16
@Thorgott yes, but what's the optimal algebra example?
"optimal" as in "minimal"?
02:55
Guys , in Microsoft word 2007. How to reduce spaces between lines
I want to write it in a way it looks like division
anyone know how to do the above? Showing it's quadratic convergence i.e. $e_{k+1} = O(e_k^2)$
 
1 hour later…
04:08
@Thorgott yes ... or you might say ... universal.
something like $\mathbb{Z}[x,y,z,w]/(x-zy,y-wx)$?
so whenever you have $(a)=(b)$ for $a,b\in R$, $R$ some ring, you can find $r,s\in R$ s.t. $a=rb$, $b=sa$, then you send $x\mapsto a,y\mapsto b,z\mapsto r,w\mapsto s$
so this ring is universal among all rings with a choice of mutually divisible elements and mutual divisors (without the choice of $r,s$, there's enough ambiguity for me to believe you won't find something meaningfully universal)
since in some rings, $a,b$ don't differ by a unit, $x,y$ don't do so in this universal ring (or just see directly that the only units in this ring are $1,-1$)
04:23
Hmm, I did this (admittedly in a $K$-algebra) in a ring with three generators, not four.
Interesting. Presumably yours maps to mine but not vice-versa (not worrying about $K$ versus $\Bbb Z$).
I mean, $\mathbb{Z}[x,y,z]/(x-zy,y-zx)$ works to produce some counter-example, but I think it's not universal in a relevant manner anymore
04:55
Hi
I have stopped talking for no reason
05:45
apparently not
06:16
@LeonhardEuler thanks
 
3 hours later…
09:26
@Thorgott A "purified" version of the C(R,R) example : consider the subring of $\Bbb Z_2^3$ generated by (1,1,1), (1,0,1), (1,0,-1)
09:45
@Astyx now what
10:03
I meant $\Bbb Z_3$
10:17
Does anybody know an application of probability in pure mathematics, e.g. something like probability used to calculate fundamental groups/homology groups?
11:02
@Proxava there is Brownian motion on Riemannian manifolds but I don't know anything about it
also there's a statistical proof of RH somewhere
Riemann Hypothesis
11:32
thank you, in the mean time i found this: math.stackexchange.com/questions/4983/…
but it is a bit disappointing, I heard of BM on manifolds, but do you maybe have a good reference?
 
1 hour later…
12:32
For s>1/2, the above is equivalent to the Riemann hypothesis.
$$\displaystyle \frac{1}{\lim\limits_{x\to \infty } \, \left(\sum\limits_{a=1}^{x} \frac{1}{a^s}+\frac{1}{(s-1) x^{s-1}}\right)}$$
$$= \lim_{x \rightarrow \infty} \left( 1 - \sum_{2 \leq a \leq x} \frac{1}{a^{s}} + \underset{ab \leq x}{\sum_{a \geq 2} \sum_{b \geq 2}} \frac{1}{(ab)^{s}} - \underset{abc \leq x}{\sum_{a \geq 2} \sum_{b \geq 2} \sum_{c \geq 2}} \frac{1}{(abc)^{s}} + \underset{abcd \leq x}{\sum_{a \geq 2} \sum_{b \geq 2} \sum_{c \geq 2} \sum_{d \geq 2}} \frac{1}{(abcd)^{s}} - \cdots \right)$$
for $s>\frac{1}{2}$
 
1 hour later…
14:00
Hi Astyx.
14:24
Hi
14:58
@Proxava there's a nice proof of the fundamental theorem of algebra using martingale convergence. But fundamental groups/homology? Never heard of an example, but that might be my ignorance
what's left implicit here is that every bounded local martingale is a bounded martingale and hence converges almost surely for t -> infinity
Ofcourse the instructors are making a mess of Linear Algebra @Ted, they are using Axler lol
15:24
@user2103480 Yes, this is usually done in a first course
There is a degree argument and a winding number argument which are basically the same.
The winding number argument: if p(z) has no zeroes, consider the winding number of the loop p(S(r)), where S(r) is the circle of radius r, in C - 0. Clearly p(S(0)) is constant so has winding number 0, and also because p has no zeroes, p(S(0)) is homotopic to p(S(r)) in C - 0 for any r.
But for large r (larger than, say, the sum of absolute values of coefficients) this loop is explicitly seen to be homotopic to r^n z^n.
Contradiction as soon as you know the winding number is a homotopy invariant.
@user2103480 Do you need Ito's formula to see $f(B_t)$ is a martingale? The real and imaginary parts of $f$ are harmonic, so doesn't that follow from the mean value property?
Very nice proof
15:52
@user2103480 Sorry I didn't read up before and misread your comment. You probably know what I said above.
16:46
@Astyx ah, nice
@SayanChattopadhyay Axler is definitely meant for an upper-level course, not for students who've never learned basic linear algebra. Who knows what they're doing ...
Let $T$ be a 3-regular tree. Does $\text{Aut}(T)$ act transitively on the Cantor set of ends of $T$?
@Ted this is so stupid. I was discussing vector spaces with a few students and the only example of a vector space they know is the set of all functions from natural numbers to R. This is horrendous teaching
17:04
They'll compensate
@BalarkaSen How do you answer questions like this?
Is there more than one 3-regular tree?
Seems like no
I think that gives a proof actually.
The usual picture of a tree
What is the proof?
Just heuristic, but the ends correspond precisely to choosing an infinite sequence of (never-repeating) adjacent nodes, right?
Yes, but that only gives topological transitivity. Switching branch gets you arbitrarily close
And you get a cantor set by identifying the ends with $3 \cdot 2^\Bbb N$
I was going to suggest disassembling and reassembling the tree.
You can map an end $x$ arbitrarily close to any $y$ in the end by switching branches
17:07
Inductively
I dont see why you can't do an infinite disassembly
You must give me an automorphism of the tree. I don't see how to do that by disassembling everything
You clearly understand my idea at least as well as I do though so I trust that you have caught some subtlety
No I could be wrong
I am just a regular idiot
Basically I think I know what to do but I struggle with notation
You already know what to do if you just want to swap two branches.
You want to take a limit of those automorphisms maybe
17:09
As you go down the branches, indeed, this gives you an automorphism moving around the first n digits of the end
Which I agree is not enough
But the n'th automorphism fixes the subtree of things reachable from the center by n steps
So I think a limiting procedure works fine
That makes sense
Yah OK so you had the proof, and what neither of us realized is that there is a natural well behaved notion of convergence here
Weak convergence of automorphisms I guess
Ye, Aut(T) is a topological group in a clear sense
Now writing this down sounds like a mess lol
17:11
You want some sort of digit notation for nodes on the tree
I can just write down the automorphism explicitly.
Sure, I just struggle with what explicit means here since I am not used to how people talk about trees.
The description above is clear to me but maybe not to a probabilist
I was going to use your notation haha
I suggest digit notation with obvious notions of adjacency
Write ends as infinite binary sequences. First time they disagree, switch branches at those corresponding levels. Repeat. To infinity.
17:13
Because then it will be clear how to write down a formula, what digit is sent to what digit, and it will be clear that adjacent guys are sent to adjacent guys
This is a well-defined automorphism of the tree of course
Nice, thanks, this is very clean.
Yeah so a node is an integer 0,1,2 and then a finite sequence of 0,2's (other than the root, which is the empty string). And two node are adjacent iff the strings differ only in the last slot
If handed an infinite binary sequence one ought to then be able to write down a formula
Algebra brain
Lol
I was getting confused precisely because of algebra. Cayley graph of $F_2$ is a 4-regular tree, but of course $F_2$ doesn't act transitively on the ends, only topologically transitively.
$F_2$ is a countable group. But $\text{Aut}(T)$ has infinite words
No problem
Good stuff. Thanks!
17:41
So for $A$ a local ring, $M$ its maximal ideal, $a \in A \setminus M$, $m \in M$, how come $a + m$ a unit? $a$ is a unit and $m$ a nonunit because of Zorn's Lemma and $A$ being local. That much I have figured
So how do you show $a$ must be a unit?
since $b$ a nonunit must be in a maximal ideal and there is only one, all the nonunits are in $M$, so whatever remains is a unit
So doesn't that verbatim answer for $a+m$?
You should not have to think at that
17:46
G'day, a @Balarka.
hmm I guess $a+m$ has to be in $A$, but if it were in $M$, then $a + m - m \in M$?
Forgive me for being rude, but ... duh.
hahaha no stress. it is really obvious I guess
silly that I did the "hard" part and then got stuck on this easy bit
We all can be stooopid sometimes. But sometimes this chatroom does scare me.
17:50
you are very brave to come back here day after day
18:00
Meanwhile, I'm getting ruder to no-effort people.
yes, the point is that the maximal ideal in a local ring consists precisely of the non-units
this also works the other way: if the non-units in a ring form an ideal at all, the ring is necessarily local and the maximal ideal is the ideal consisting of the non-units
@Thor: Have any further thoughts about my non-associate ring question?
Oh, oh. Fargle is awake early today.
Much to my lack of amusement.
no, I believe what I had is the closest to something "universal" one could reasonably get
did you see the example Astyx gave earlier? I think that's probably minimal
Well, my answer is $K[x,y,z]/(z(1-xy))$.
No, I didn't see Astyx's.
18:06
he algebra-ified the C(R;R) example: consider the subring generated by $(1,0,1),(1,0,-1),(1,1,1)$ in $\mathbb{Z}_3^3$
Well, @Fargle, we can try to amuse you, but it won't be pretty.
Oh, that's interesting, @Thor, but far from "universal" in the sense I was thinking.
And @BigSocks still needs to prove that it in any quotient of a PID, $(a)=(b)$ iff $a,b$ are associates.
@TedShifrin oh, I see what you're doing
Yeah. I'm taking the proof that one normally does :)
right, that supersedes the four-generator example I gave
In fact, when this question arose, I did talk with Lorenzini (whom BigSocks is reading) and he concurred that my example was optimal. :)
18:09
because once you have two elements dividing each other and you have two divisors, you can express each as multiple of the other
The hard part is to show that the units of this ring are just $K$.
For some reason, I've found this a slightly intriguing question.
@TedShifrin You made me have to double-take---my brain inserted a comma between the $z$ and the $1 - xy$, so I was perplexed for a moment.
You might want to smack your brain for confuzling you.
I'm no brainologist but there's no way that's good for it.
@TedShifrin that just comes down to saying $z-xyz+1$ is irreducible, which isn't too bad
18:15
Yeah, I think that's how I thought about it. But this was 20 years ago.
@TedShifrin Seems easier to just not talk to them. About 50% of why I deleted my account.
Change your $\Bbb Z$ to $K$ and then your example should surject to mine?
Well, it's my tiny bit of ethical police-work, @MikeM. Probably for nought.
yes
but if you want universality, you should change the $K$ to a $\mathbb{Z}$ instead
Right, I agree.
Well, they will never learn what they are purportedly trying to learn, which is at least minor punishment.
18:20
I just stumbled on a doctoral student who can't look up the definition of laplacian and compute the harmonic polynomials of fixed degree on $\Bbb C^3$. AGH.
Oh oh, @robjohn is here with more decorations for us.
As time passes I find myself not able to do more and more basic math
You're already a grad student, a @Balarka :P
I can still balance a checkbook, which is stunning.
I was, in 11th grade.
I have the opposite problem, @Balarka---I find myself desperately clinging to the basics as all the rest falls apart, haha
I dropped out and forgot all math
18:22
Hippa should make a Balarka meme for us. The hell with polar coordinates.
@Fargle Top-down or bottom-down, that's the difference
Everything falls apart anyway
@TedShifrin I've improved my Easter avatar, but I haven't revealed that one yet :-)
You have months left to work on it.
@TedShifrin And I may... recently I've written new and improved code for Mathematica to make better graphics. I spent three days improving my text along a curve code for one answer.
I have not monkeyed much with Mathematica in over 10 years, although I did buy the latest. I bitched to you about how Wolfram told us we needed 12.1 to run on Big Sur. So I gave in and bought it. Guess what! 12.0 still opened up when I tried a week ago. Grr.
18:33
I need to upgrade my OS to run 12.1, which I have but sits dormant on my computer for now.
this is the graphic I used in an answer that was three days in the making.
Holy cow! You must have spent hours and hours on that.
I used to draw stuff in Mathematica, but add the labels using LaTeXiT in Illustrator (which, of course, I no longer have).
@TedShifrin That sounds hard to me
Just a linear algebra calculation, @MikeM. Tedious, but why hard?
Cubic polynomials.
@TedShifrin The long part was getting the code to fill the weird shaped curves (defined by complex functions) and to put the text along them. The one problem is that the long horizontal bars are just lines and don't get curved by mapping the control points.
Unless you mean of like degree 5
In which case sure it is a linear algebra problem.
Oh ok.
18:41
Plus you did the colors and shadings of the text. Way more intricate than I have patience for.
I thought you meant "of fixed degree d".
Yeah, just cubics, @MikeM. Sorry, my rant was vague.
Yes, I agree that the general case would be very hard to nail down. I don't know if there's an inductive approach.
There is so much to teach in one linear algebra course. :(
I guess a lot of teaching happens in exercises.
Well, particularly proof-learning, but you're not doing so much of that. I always wanted to be sure to do enough applications (and included both spectral theorem and matrix exponentials and ODE at the very end). Most of my colleagues never got anywhere close.
I don't know what's become of the new applied LA course that started as I left UGA.
I do a little (basic) proofs. Maybe 20% of the content is proof based.
18:45
I stole from Strang the idea of doing a section on graph theory, relating topology to basics physics notions. But I never had time to teach that in class.
But it was fun to write.
I worry I won't have time for orthogonal matrices and GS, much less the spectral theorem. But I am peppering in things in supplementals as we go. I had one about adjacency matrices and path counting which will lead well into Markov chains later (once you weight the edges).
QR algorithm is big for applied people, I think.
I like the path counting because it is a visual way to see matrix multiplication.
Or numerical people, I should say.
Yes, I have the path counting as an example, too. I saw that in the very first book I taught out of in 1982.
I just wish it were two sems long to be honest.
18:46
You just can't do everything :P
There's enough content in linear algebra for eight semesters.
LOL ... well, so much of graduate differential geometry is (multi)linear algebra.
But I probably can't sell my eight semester sequence to my chair.
Um, no, probably not.
I could manage two; I don't think I can manage eight.
Fun fact: There's a way to interpret the rational canonical form in terms of Markov chains
18:50
You mean the companion matrix?
The one with those, yeah
A semester on quadratic algebra. A semester on cubic algebra. A semester on derived linear algebra. ...
A semester on cubic algebra? Oy.
Normal forms for cubic forms may not even be known?
Guys, why is $g$ equal to this quotient? So far they only argued it's equal to it locally, no?
If $X$ were irreducible, then we would have the equality automatically
but in this general case I don't see it
@TedShifrin It will be a research seminar.
18:53
The most important cubic form is $x_1^3 + a x_1 + Q(x_2, \cdots, x_n)$ where $Q$ is a quadratic form.
(Weierstrass form)
LOL, OK.
I jest, obviously. But I think two semesters is honestly the right length.
Yes, I agree that I could fill two semesters with an excellent course.
Especially if we're doing numerical stuff.
I even have a section on projective geometry and computer graphics in our linear algebra book. Never got to teach that, either.
But people interested in computer graphics have read it.
I try to emphasize the computer graphics POV occasionally.
I'm sure they love moving up to $\Bbb R^4$. :)

« first day (3832 days earlier)      last day (1487 days later) »