« first day (4361 days earlier)      last day (955 days later) » 

00:32
@Thorgott May proved $\pi_i(S^n) = 0$ for $n>i$ using smooth (or simplicial) approximation and various approximation theorem appears in the later part of the book.
It's in chapter 9 section 4.
01:16
what's that got to do with the question?
Oh you're talking about colimit thing. He assumed topological spaces are all compactly generated in the whole textbook. I thought you're answering my latest chat sry
compactly generated or compactly generated hausdorff?
He didn't assume hausdorffness but he assumed compact spaces are always hausdorff. I don't know this is equivalent to compactly generated hausdorff
Does anyone know the name of approximation theorem the proof? Is it Whitney?
01:37
yeah ok, he uses "compactly generated" to include "weak hausdorff"
and that makes the lemma in the question work
but it's very untrue without any sort of separation hypothesis
@onepotatotwopotato Whitney approximation, simplicial approximation, CW approximation, they all do the job
How Whitney approximation is applied explicitly?
it tells you the map is homotopic to a smooth one
smooth surjections cannot increase dimension, so it then has to miss a point
the latter fact requires a bit of work if you don't know it
Oh I see thanks.
01:52
Go Sard.
One more. The proof uses smooth approximation again but this time, homotopy is relative to $\partial I^q$. How this is possible? There's no assumption about smoothness of $f$ on $\partial I^q$
And how can we control the missing point to be in the interior of $D^r$? We only know the missing point exists somewhere
 
1 hour later…
03:11
They coglue, obviously. — Najib Idrissi Jan 8, 2015 at 13:19
 
2 hours later…
05:36
Quiet in the summer time.........
05:47
🤫
 
2 hours later…
07:18
@Thorgott IMO, the more analytic arguments enrich this hyperbolic geometry argument. The cover $\lambda : \Bbb D \to \Bbb C \setminus \{0, 1\}$ is a modular form.
The way they construct this analytically can also be conceptually understood as follows. Use the Poincare upper-half plane model $\Bbb H$, then points $\tau \in \Bbb H$ here are in 1-1 correspondence with elliptic curves $\Bbb C/(\Bbb Z + \tau \Bbb Z)$ upto isomorphism.
Any elliptic curve $E_\tau := \Bbb C/(\Bbb Z + \tau \Bbb Z)$ admits a branched 2-fold cover to $\Bbb{CP}^1$, namely the Weierstrass $\wp$-function $E_\tau \to \Bbb{CP}^1$. Topologically, this is like skewering an axis through a torus which cuts it at 4 points, and then doing a $\pi$ rotation along it. The quotient is a pillowcase, which is $\Bbb{CP}^1$ with four marked points $z_0, z_1, z_2, z_3$ which are the branch points of this cover.
Then $\lambda : \Bbb H \to \Bbb C \setminus \{0, 1\}$ is defined by letting $\lambda(\tau)$ to be the cross-ratio of $z_0, z_1, z_2, z_3$.
$0, 1$ are not in the image because all those points are distinct.
$\infty$ also, of course.
@Thorgott Also correct. The deck transformation group is in fact $\Gamma(2)$, the kernel of the mod 2 reduction $\mathrm{SL}_2(\Bbb Z) \to \mathrm{SL}_2(\Bbb Z_2)$. Number theorists would say this means $\lambda$ is a "weight $2$ modular form"
This is related to the fact that one of the branch points $z_0 = \infty$ always, and there's a $S_3$-symmetry permuting the rest $z_1, z_2, z_3$. Indeed, $S_3 \cong \mathrm{SL}_2(\Bbb Z_2)$.
Not weight, level is the right terminology.
 
3 hours later…
10:58
I tried to prove this math.stackexchange.com/questions/4491637/… using this: let $x_0 \in \mathbb{R}$. It is $f(x_0)=f(x_0+1)=\dots=f(x_0+n)$ for any $n \in \mathbb{N}$. Since $f(x) \to L$ as $x \to \infty$, for any $\epsilon>0$ there exists $M_\epsilon>0$ such that $x>M_\epsilon \implies |f(x)-L|<\epsilon$. So, if $n>M_\epsilon-x_0$, it is $|f(x_0)-L|<\epsilon$ for any $\epsilon>0$, that is $f(x_0)=L$.
I am not sure this works because $M$ depends on $\epsilon$, from $n>M_\epsilon-x_0$ it should be $n=n_\epsilon$ too and so I am not sure if this dependence on $\epsilon$ can make the proof invalid.
11:10
That's OK. what you showed: for all $x_0\in \Bbb R$, for all $\epsilon > 0$, $|f(x_0) - L| < \epsilon$. How you showed this, relied on a point that was very far away, that depends on $\epsilon$ and also $x_0$. But that point always exists, for each fixed $x_0$, and each fixed $\epsilon$ @Gwyn
@CalvinKhor thank you for the answer. You mean that I'm relying on the point $M_\epsilon-x_0$ and that this point always exists because, fixed $\epsilon$ and $x_0$, it is $M_\epsilon-x_0$ fixed and since $n$ can be taken arbitrarily large I can always be sure that $n>M_\epsilon-x_0$ is satisfied?
11:50
Please would someone help me repair the following answer?
1
A: Show $S_7$ is isomorphic to the subgroup of all those elements of $S_8$ which leave the number $5$ fixed

ShaunI'm sure this has been answered here before but, since I can't find a duplicate, I'll give a CW answer. There's nothing special about $7$, $8$, and $5$ here. Theorem: Let $n\in \Bbb N$ and $m\in\{1,\dots, n+1\}$. Then $S_n$ is isomorphic to the subgroup $H$ of $S_{n+1}$ whose elements fix $m$. ...

It's CW and I have already made an edit, although I still don’t think it's right.
@Shaun I'd rather re-indexing first and assume $m$ to be $n+1$.
Ah, I see! Thank you, @onepotatotwopotato.
@Gwyn yes, though it should convince you without me!
12:05
maths
12:15
is not a spectator sport
13:13
@BalarkaSen interesting. not visually obvious to me.
 
3 hours later…
15:50
Let $f$ be integrable on $[a,b]$. Suppose $f(x) \ge 0$ for all $x \in [a,b]$ and $\displaystyle\lim_{x\to x_0} f(x) = f(x_0)$ with $f(x_0) > 0$ for some $x_0 \in [a,b]$. Prove that $\int_a^b f > 0$.

Since $f$ is continuous at $x_0$, we know that for any given $\varepsilon > 0$, there exists a $\delta > 0$ such that $|f(x) - f(x_0)| < \varepsilon$ whenever $x \in [x_0 - \delta/2,x + \delta/2]$. Therefore, set $\varepsilon = f(x_0)/2 > 0$ and notice that $-f(x_0)/2 < f(x) - f(x_0)$ so $f(x_0)/2 < f(x)$ for $x$ in the aforementioned interval. Thus, $f(x_0)/2$ is a lower bound of the set $\{ f
Does this quick proof work?
i didn't skim the details but that is definitely the idea
Well, the details is usually what gets me :D
fix an epsilon for which you have some bound of the form f(x) > p > 0 for x in an interval, then use a partition having that interval to compute a lower bound on the integral
I think that is exactly what I did
that's what it looks like to me. i'm not sure why you chose delta/2 instead of delta. does that matter? there's maybe also some annoying crap about what if x_0 is a or b so that f is not defined on a 'whole' interval around x_0.
15:56
Can anybody provide me with an example of a nonempty set $G$ which is closed under an associative product $\cdot: G\times G\to G$ with right identity ($a\cdot e = a$) and left inverses (There exists $y(a)$ such that $y(a)\cdot a = e$) which is not a group?
what if you take the integers (or really any nonempty set) with a * b = a for all a, b
@leslietownes Because $f(x) > f(x_0)/2$ for $x \in [x_0 - \delta/2,x_0 + \delta/2] \subseteq (x_0 - \delta,x_0 + \delta)$, but in the definition of $m_2$ we have a closed interval as well
So I choose a subset
@leslietownes Ahhhh, that's smart. Thank you for the suggestion!
Or subinterval
16:18
good morning, ted.
16:29
I don't really understand the one-point compactification of the union of a countable family of disjoint open intervals. Can anyone walk me through it?
@geocalc33 You add a point, wave your magic wand, and *PRESTO*! It's compact!
right
Understand it?
Walk you through what?
16:32
I was wrongly thinking about disjoint
What is that space then homeomorphic to?
It's not just one space though
Greetings, Munchkin’s pet.
@Jakobian but it's a topological space right?
A bunch of circles attached at a point.
I mean... of course it is
yeah. I think it can be any finite or countable wedge sum of circles
and that's all of them, depending on the amount of the open intervals
16:37
Good morning professor
Morning, polite.
Actually, the countable amount is a bit tricky, because they could get smaller and smaller
Like a Hawaiian earring
It’s not a subspace of anything.
No, definitely not that.
For finite amount of intervals it should be just this en.wikipedia.org/wiki/Rose_(topology)
For infinite amount it could be Hawaiian earring or a countable wedge sum of circles, which are already different spaces
I don't think there's any other possibility
Wrong topology to get earring.
A small neighborhood of $\infty$ should not contain entire components of the space, just complements of compact subsets.
I haven’t checked definitions in many years, though.
Oh, apologies. I think you are right. Complement of compact subset of the whole thing.
 
2 hours later…
18:18
In line for x-rays and no one is entertaining me here!
@TedShifrin I remember working out the transform for CAT scans while I was waiting in church with my parents. You could try that ;-)
@TedShifrin Even given the proper topology, I probably wouldn't get an earring.
My genus is high enough as it is.
@TedShifrin Are you okay?
@TedShifrin oh no, check up or something serious :(
Can someone tell me why we cant use the theory of L'hopital here
$ lim_{x\rightarrow 1} \frac{x^3+x^2-x-1}{x^2-1}$
@MadSpaces Why can't you?
You get as a result =4 the real result is 2.
18:33
$\frac{3+2-1}{2}=2$
Yeah, you do get $2$.
$(x^3 + x^2 - x - 1)' = (3x^2 + 2x - 1)$, so plugging in $x = 1,$ we get $3 + 2 - 1 = 4$, and $(x^2 - 1)' = 2x$, so at $x = 1$ we get $2$. So $4/2 = 2$
I would check your differentiation on the denominator; it's probably the culprit
18:54
thats not how l hospital is used
oh
nvm i am stupid
@polite @Mad Just check-up … plus finding out how bad my hips are to go with bad neck and back :)
@TedShifrin These geometry proofs using vectors sure take me back to AoPS Precalc...
(in your text)
19:23
It is easy to prove that if $f$ is $o(x^n)$ then its integral function $F$ is $o(x^{n+1})$, using Hopital's rule for the ratio $\frac{F(x)}{x^{n+1}}$. However, after differentiating numerator and denominator, there is a factor $n+1$ at the denominator that makes this proof not valid for $n=-1$.
I tried to find a counterexample for $n=-1$: for instance, $e^{-x}$ is $o(1/x)$ as $x \to \infty$ but $F(x)=\int_a^x e^{-t}dt=-e^{-x}+e^{-a}$ which tends to $e^{-a}$ as $x \to \infty$ and so $F$ is not $o(x^{-1+1})=o(x^0)=o(1)$ because $e^{-a}$ does not tend $0$ as $x \to \infty$ for any $a \in \mathbb{R}$. All I said is correct?
19:35
@politeproofs When I taught the AOPS precalculus course three or four years ago, I added a number of such problems to the course.
Yeah, you made my life harder
@Gwyn Are you doing $x\to\infty$ or $x\to 0$? In either case, I think $n>0$ is the only thing that makes sense.
@politeproofs I did?
Indeed, I always prefer the analytical method of solving a problem and drawing a picture is always sometimes I doubt myself doing
A picture isn't a proof. It's guidance.
I know, but for these proofs a good picture will basically give you the entire proof and you just have to write it down
19:38
@politeproofs Are you suggesting that you were in my AoPS class? I'm confused.
No, I wasn't, but you said you added those flavor of problems to the classes
(unless yours were removed later?) But I doubt that
My additions to my AoPS calculus and precalculus classes were just for my students. It's possible they incorporated some of my stuff into the courses afterwards. I complained a lot about some of the precalculus course.
There is a special week dedicated to geometry with vectors
Anyhow, if they changed the courses after my complaints/suggestions, they never told me.
Precalculus is pretty universally a terrible class. :/
19:40
Yes, it is. But AoPS incorporated a fair amount of complex stuff that just didn't belong there.
Subtle stuff with conformal mappings that was beyond high school level, IMHO.
To me, precalculus should be about mastering trig and exponential/log stuff (and the AoPS course had none of the latter) and lots of word problem skills for college calculus and applications to chemistry, physics, etc.
But I hated teaching courses that had the feeling, "You'll see why this is important/interesting when you take ..."
The skills you are talking about were taught to me in introduction to algebra by AoPS (the book, I never took the course)
It goes intro algebra -> intermediate algebra -> precalc -> calc
Right, the exponential stuff was in algebra, but I am never sure how well students learn it there. Certainly regular math students don't.
But we did tons of trig.
I mean I completed both precalc and calc with an A+ and I can't imagine not acing any high school test
19:44
The calculus course was very disappointing to me, especially because all of my students were simultaneously taking AP calc in high school. But I incorporated more proofs and asked them to do some. The AoPS homework was not particularly interesting.
For that purpose, the classes/books are great
Well, that shouldn't be the purpose.
I think the AoPS "view of mathematics" runs out of steam by high school level.
Well, comparing AoPS Calc homework to the AP Calc BC test makes the test look like a joke
Maybe that's too harsh, but I do think it's two completely different levels
I suggested to my boss that we could teach some really interesting stuff to the most talented students — both he and I were PhDs in geometric/topological math. But the powers that be didn't want to do anything that couldn't be plugged into any class available anywhere.
Well, the AP BC exam is horrible in the first place. And giving 5s to anyone who gets 60% or better is worthless. So many kids exempt out of two (or more) semesters of college math and should not.
You get a 5 if you get over 60%? Haha I never knew that
19:46
It's ridiculous.
I thought it would be.. ≥90%
Maybe 80%
The kids I taught for 35 years who had top-notch teachers and did well were fine, but plenty did not and were not.
Nope.
I took the BC test in 1970. I had to write a $\delta$-$\epsilon$ proof.
I had taught it all to myself, anyhow.
@TedShifrin But the first were a subset of anyone who took AoPS classes and did well
Sorry, the other way around
None of my students at UGA had anything to do with AoPS.
Yeah, I realize that
But people who did, were "well-prepared"
(or would you disagree?)
19:48
I was not impressed with the AoPS course that was fed to me to teach. But I taught what I wanted to teach. I had permission. They conceded that I was "qualified."
I haven't had an incompetent teacher there though. My first teacher had a Stanford PhD and a gold medal from the IMO. My second teacher was head of undergraduate mathematics at his university
What I taught included everything they had, but my students basically had no time to put into the course, anyhow, as they were too busy with high school and extra-curriculars. I wrote a few recommendations for one of my students, but none of the rest of them has stayed in touch to let me know how college has gone.
Another teacher was a graduate student, but he was very competent
You did it all on-line?
Yes
I am not in the US
19:50
I figured that. ... I actually taught in person to my handfuls of students.
I know, I am jealous of them
The in-classrooom technology malfunctioned lots. Very infuriating.
I sometimes did most of the class on a tiny, tiny whiteboard. Also very infuriating.
Anyhow, I miss teaching, but I'm glad I'm a bum now.
Anyhow, the last of the geometry vector proofs you mentioned in my book are quite interesting. Not stuff most of us in the US know.
But obviously you shouldn't be worrying about that stuff.
All your upper- and lower-sum stuff will return in Chapter 7.
Could always make videos ;) pen-and-paper style at your own pace. Richard Bocherds does that, and I'd love to watch his videos but unfortunately they're mostly for graduate students
There are (very few) pen-and-paper videos of undergraduate mathematics available
And I'm not talking about non rigorous calculus 1-2
19:53
I felt that having actual live students in my class, participating, made those videos a lot better than artificial pen-and-paper teaching would be.
I mean multivariable analysis lite, topology, abstract algebra based linear algebra (but undergrad not grad)
I sort of wish I had convinced the students to videotape my last differential geometry class, but they were exhausted from filming.
You can't do abstract-algebra based linear algebra other than in the context of a good algebra course. Read Artin. :)
Oh oh. It's a @Balarka.
Hi @Ted
I mean you can introduce things like fields, groups, semi-groups, and so on without going to Artin I think.
Axler alllllmost goes there, but not quite.
You'd be happy to know that I am reading Arnol'd
19:56
I really dislike Axler's book/course, so that doesn't sell me on anything.
The point of having abstract algebra is to use module theory to do canonical forms.
Which, Balarka?
Doing linear algebra over finite fields is more motivated in algebra and its applications.
OK, lunchtime for me.
Singularities of Differentiable Maps, vol 2
It has a nice introduction to Lefschetz fibrations
@TedShifrin youtube.com/watch?v=TCXNhBlBNCg this is what I used for Linear Algebra. It's a course by the University of Toronto during the pandemic, and it uses Axler but the professor does really well to go over abstract algebra concepts and the course goes further than the book in a lot of aspects
It's honestly the best set of videos on Linear Algebra at that level that I could find by far, nothing comes close
The only issue is I had to sit on the student-made discord server, and non-suspiciously ask people for a screenshot of their homework, and it didn't always work. So I had a lack of very interesting homework problems (because the ones that I was able to get were very informative to complete)
Never read that, @Balarka.
Bubye all for now.
@TedShifrin Actually I was interested in both cases, but the example I found is for $x \to \infty$. Thank you for your help, I agree to that this works only for $n>0$.

« first day (4361 days earlier)      last day (955 days later) »