« first day (3651 days earlier)      last day (1666 days later) » 

00:24
@Calvin: There's nothing wrong with starting with integrals. The estimate ideas are still already present.
@TedShifrin did you know that posets also have homotopy theory... lol
what nonsense
@TedShifrin you don't think the limit definition is easier than the definition of Riemann integrability?
Sure it can be done, just saying that starting with easier things might make things easier
 
1 hour later…
01:51
@CalvinKhor No, I don't. Archimedes had the theory of the integral down many centuries (count them) before Cauchy thought about limits. The quantifiers in the limit definition are far more difficult.
02:04
pretty sure Archimedes didn't exactly worry that the mesh size was less than delta to ensure the difference of upper sum and lower sum was less than epsilon @TedShifrin
02:50
My whole point is that the quantifiers are simpler with integration. But you should actually read some of the history before being so sure of yourself. They totally understood that you had to squeeze from both sides with the difference of outer and inner estimates going to $0$.
 
2 hours later…
05:08
@robjohn @CalvinKhor I think I’m missing something substantial in my understanding of limits. As far as I know till now, we are given (or assume that we are given) the epsilon and we need to find the delta.
So, if I were to do the proof of what you gave yesterday, I might do this:
Suppose $\lim_{x \to a} =L$. Then, by the definition, of the limits, for a given $\epsilon \gt 0$ such that $$|f(x)-L| \lt \epsilon$$ whenever $$|x-a| \lt \delta$$. There exists $x_1 \in \mathbb Q$ and $x_2 \in \mathbb T$ within $\delta$ of $a$.
So, for those values we have $$ |x_1 -L| \lt \epsilon \\ |L| \lt \epsilon$$ but we know $$|x_1 -a| \lt \delta $$
$$|x_1| \lt |x_1 -L| +|L| \ 2\epsilon \\ |x_1| \gt |a| - |x_1 -a| \gt |a| - \delta \\ |x_1| \lt 2\epsilon \\ |x_1| \gt |a| - \delta$$
Now, it $|a| -\delta$ is less than $2\epsilon$ then there is no contradiction at all.
This is what my first thought says, now how should I think so that my it would become inescapable to take $\epsilon = \frac{|a|}{4}$
and the clever way of taking $|x_1|\lt |a|/4$?
‘We are given and we need to find’ is a good slogan for proving existence of limits
In this case you are proceeding by contradiction so you assume that the limit exists ie that ‘if you gave me an epsilon, then I guarantee that I can give you a delta’
The rest of it looks like rob’s proof so I’ll leave that alone
 
2 hours later…
07:30
@CalvinKhor How we got the idea that we should take $\epsilon = |a|/4$ ?
You wouldn’t, because once you find one epsilon that works, you can use any smaller one
07:55
@CalvinKhor Why that particular choice for $\epsilon$?
How you got that idea? (Imagine you're doing it for the first time)
08:51
For the Sieve of Eratosthenes I've seen optimizations done by applying wheel factorization. What exactly is meant by this? The Wikipedia article has some info which doesn't make sense to me. Usually wheel factorization is the next level after sieving, so I don't really get this hybrid approach. en.wikipedia.org/wiki/…
09:12
@Knight I "did it for the first time" with the sequential formulation of a limit, which as i said already imo is much more intuitive and easy to write. The only thing that I think is useful to say about the above proof attempt, is that you did not take advantage of the freedom to choose $x_{i}$ to be arbitrarily close to $a$. The $\delta$ given by the existence of the limit could be really large. But if it is, you can always use e.g. $\delta/2$ instead.
Oh and draw the picture, with the delta neighbourhood and the epsilon neighbourhood too
Hello, I'm studying in a summer camp, where the prof. is gonna talk about the de Rham cohomology two weeks later. So, I'm previewing some concepts involved, but I couldn't find any proof/argument that deals with the relationship between simplicial/singular cohomology and De Rham cohomology. Is there any reference? Thanks in advance. (Sorry, I don't know if asking this in this chatroom is appropriate, maybe I should write a question)
 
1 hour later…
10:28
hey I'm her here
I'm new here :D *
11:19
Can anyone point to any resources of how to represent Queues using Differential Equations?
11:43
What is wrong in my answer below? It looks weird but correct?
0
A: Help with solving for $a$ in $\ln(x-a) = \frac12 \ln(x-b) + \frac12 \ln(x+b)$

Artificial StupidityAnother trick or threat. $$ 2\ln(x-a) = \ln(x-b) + \ln(x+b) $$ Applying $\frac{\mathrm{d}}{\mathrm{d}x}$ to both sides, we have \begin{align} \frac{2}{x-a} &= \frac{1}{x-b} + \frac{1}{x+b}\\ \frac{2}{x-a} &= \frac{2x}{x^2-b^2}\\ x^2-b^2 &= x^2-ax\\ a &=\frac{b^2}{x} \end{align}

one differentiates functions, not equations
 
2 hours later…
13:27
is the self intersection of two transverse lorentzian manifolds of dimension $D=3+1$ also lorentzian?
the self-intersection of two manifolds?
actually I mean the intersection, not self intersection
does it just follow from the fact that two transversal manifolds is a manifold?
and since you're talking about transversality, I assume those two manifolds are submanifolds of an ambient manifold
is the ambient manifold itself Lorentzian and your two submanifolds are Lorentzian submanifolds?
13:44
yeah the ambient manifold is lorentzian and the two submanifolds are lorentzian
14:07
I think the answer is "yes". Let $M$ be a Lorentzian manifold and $M_1,M_2$ Lorentzian submanifolds that intersect transversely. The transversality theorem implies $N=M_1\cap M_2$ is a submanifold of $M$. Let $p\in N$ and $g$ be the Lorentzian metric on $T_pM$. There is a unique one-dimensional subspace $W$ of $T_pN$, restricted to which $g$ becomes negative-definite by signature conditions, so since the same holds for $g\vert_{M_1}$ and $g\vert_{M_2}$, we must have $W\subseteq T_pM_1\cap T_pM_2=T_pN$, so $g\vert_N$ has the right signature.
@Knight You can just plug in any $\epsilon$ and see when it gives a contradiction. Plugging in $|a|/3$, just barely puts the inequalities on their boundaries, and then things will depend on whether you have $\lt$ or $\le$, so I chose $|a|/4$ to make sure the inequalities are contradicted.
15:06
@robjohn Sir, can you please share the deep thinking behind this statement of yours $$|x_1-a| \le min ( \delta, |a|/4)$$?
 
1 hour later…
16:09
yes deep thinking please share the deep state of thought you had at the point of comment
16:24
We are given $$a^2 +b ^2 =1 \\ x^2 +y^2 =1 $$ and we need to show that $$ax + by \lt 1$$
By GM < AM we have $$ \sqrt{(a^2 +b^2) (x^2 +y^2) } \lt \frac{(a^2 +b^2) +(x^2 +y^2)}{2} \\ a^2x^2 +a^2y^2 +b^2x^2 +b^2y^2 \lt 1$$
We can leave out the two middle terms and our inequality will become even better $$ a^2x^2 +b^2y^2 \lt 1$$ Now, we know that $$(a^2x^2 + b^2y^2 = (ax +by)^2 - 2axby \lt 1$$
16:39
I think I got it, we have from GM < AM $$2\sqrt{a^2 x^2 b^2 y^2 } \le a^2y^2 +b^2x^2 \\ 2 axby \le a^2 y^2 +b^2 x^2$$
And we already proved that $$a^2x^2 +a^2y^2 +b^2x^2 +b^2y^2 \le 1 $$ Now, replacing the two middle terms by something smaller will make our inequality even stronger, therefore $$ a^2x^2 + b^2y^2 + 2axby \le 1 \\ (ax +by)^2 \le 1 \\ |ax +by| \le 1 $$
@Knight You mean: we can pick $x_1\in\mathbb{Q}$ so that $|x_1-a|\le\min(\delta,|a|/4)$? The rationals are dense in $\mathbb{R}$, so there is one in $[a-\min(\delta,|a|/4),a+\min(\delta,|a|/4)]$.
16:55
@robjohn Yes sir. Thank you so much.
 
2 hours later…
19:13
Is there an obvious reason why the figure eight is not the orbit of a point in the plane under any $\Bbb R$-action? A book I'm reading mentions this but it's not clear whether it's supposed to be straightforward
19:38
huh, interesting
@AlessandroCodenotti I assume there is some sort of continuity/differentiability assumption on the action?
right, doesn't work otherwise
20:13
Yes it's a continuous action, forgot to mention it, sorry
20:51
Solving a quadratic and I get to this form: -12 +- 2√39 / -6

Why does this simplify to: -6√39 / -3 ?

I don't understand why we seemingly divided our 12 and -6 by 2.
21:37
@Alessandro: I presume it's related to the fact that the complement has three components, but I don't see it yet.
@Beliod It most certainly does not. Perhaps you should start with the original problem? (And parentheses really do matter.)
I'm not the one that simplified this. It's on Khanacademy.
I'm assuming that I'm writing something down incorrectly. Or maybe explaining it badly.
I should have said -6 +- √39 / -3
@TedShifrin Does that change it?
22:05
Consider the map $\mathbb{R}\mapsto\mathbb{R}^2,g\mapsto g.x$, where $x$ is any point on the figure eight. The image of this map is orbit of $x$, i.e. the figure eight. Consider the stabilisator of $x$, which is a subgroup of $\mathbb{R}$, so it is either dense or a multiple of $\mathbb{Z}$. It can't be dense, for otherwise the action would be trivial by continuity, but the orbit isn't a singleton.
If the stabilistaor is a non-zero multiple of $\mathbb{Z}$, the map descends to a continuous bijection from $\mathbb{R}/r\mathbb{Z}$ to the figure eight. But the quotient is compact, so this is a
@Thorgott we say stabiliser in English btw
also nice
oops
what was i thinking
22:21
ok, the actual issue has to be subtler
cause $\mathbb{R}$ can continuously biject onto the figure eight
22:58
@Beliod No. There is a mistake there. Why don't you give the actual problem and we can settle this?
Okay.
We start off with the quadratic: -3x^2 + 12x + 1 = 0
@Thorgott Correct. I suspect it's the topology of the complement, as I suggested.
@Beliod: OK. I would recommend first multiplying by $-1$, but it's not a huge deal.
So the quadratic formula gives
$$x=\frac{-12\pm \sqrt{144-4(-3)(1)}}{-6} = \frac{-12\pm\sqrt{156}}{-6} = \frac{6\pm\sqrt{39}}{3}.$$
Yeah, since $\mathbb{R}$ is connected, the orbits will be too and since they're disjoint, the orbits of points inside either of the balls bounded by the figure eight will be entirely contained in that respective ball. So as you start to nudge a point on the figure eight just a wee bit towards the inside, one half of the figure will have to "collapse" in some sense, which contradicts my intuitive idea that the orbits should in some sense vary continuously, but I don't know if this is precise.
There's no way to simplify that other than to multiply numerator and denominator by the conjugate, but that's not something I would do. But if you do that you end up with $$\frac{1}{\pm\sqrt{39}-6}.$$
Okay, so I see where my confusion is. I probably just don't know my radicals very well (I'm trying to get better at math, so forgive the really basic questions).

When we simplify sqrt 156, we get 2sqrt39 So, what happened with the 2? Or am I misunderstanding?
23:05
That part was fine, @Beliod. That's how the 6 in the denominator turned into 3.
Factor a 2 out of numerator and denominator.
So, any time in a quadratic equation we have a number in front of our sqrt, we can use it to factor the numerator and denominator? If they're divisible by that number?
@Beliod: This is just $\dfrac{ab+ac}{ad} = \dfrac{a(b+c)}{ad} = \dfrac{b+c}d$.
Okay, thank you so much! That helps a lot <3
Okey dokey.
@Thor: It has to be continuity at $\Bbb R\times$ the vertex point of the figure 8 that's the issue.
But I'm not going to think about it no more.
23:26
Why we have $H_n(X) \approx H_{n+1}(X \times I / X \times \partial I)$ ?
Did you try writing down the exact sequence of homology of the pair?
nevermind; yes, of coures it follows from the LES
I have a tautological, and possibly unrelated question
Suppose you have a map $\sigma$ from a $k$-simplex to a space $X$. Now, partitioning the $k$-simplex to $n$ other $k$-simplices, we can restrict the $\sigma$ onto each of the simplices to get $n$ maps $\sigma_1, \cdots, \sigma_n$. The formal sums $\sigma$ and $\sigma_1 + \cdots + \sigma_n$ are formally different, but how to show the belong to same homology class, i.e $\sigma - \sigma_1 - \cdots - \sigma_n$ is homologous to $0$ ?
Because you're applying $\sigma$ to the homology of the simplex to the sum of the subsimplices.
@TedShifrin hey Ted!
hi @Stan
23:35
I was able to implement a dry run physics simulation of the linkage system! :) Very happy. Its only a basic model, but its a great step forwards
@TedShifrin so i'm having a slight problem
@TedShifrin Sorry if I was not clear, so my question is roughly equivalent to this: Let $\sigma$ be a map from a $k$-simplex to a space $X$, let $\Psi$ be a homeomorphism of the $k$-simplex. Then is $\sigma - \sigma \circ \Psi = 0$ ?
Of course you are.
So you know, with this linkage system, the vectors all pointed tip to tail
but i have something i want to draw but i'm not sure how to represent it
let me show u
23:37
I'm bit confused when something equates to 0
That's a very different question, @Lelouch.
Your first question was more like writing $[a,b]=[a,c]\cup [c,b]$.
@Lelouch You mean to ask whether $\sigma - \sigma \circ \Psi$ is a boundary chain.
But homeomorphisms needn't be homologous to the identity
Heya a @Balarka. Glad you can take over :P]
@BalarkaSen No, I mean as a chain when the (formally) equate to zero.
Not as part of the homology group
Of course it doesn't formally equate to zero!
23:39
@TedShifrin so this is what i have right now
the purple line f is supposed to be the spine
$\sigma$ and $\sigma \circ \Psi$ are distinct singular simplices, hence linearly independent in the free abelian group of all singular simplices.
@BalarkaSen Wait, so why $\sigma - \sigma = 0$ ?
@Lelouch Those are the only relations between the basis elements in a free abelian group.
so basically, $\sigma_1 \neq \sigma_2$ whenever there's a point $p$ such that $\sigma_1(p) \neq \sigma_2(p)$, right ?
even though their image may be same
23:40
Yes, that's what being distinct as simplices mean.
oh ok, thanks !
@BalarkaSen How's things going btw ?
@TedShifrin the issue is, I want to put a vector like this, but as you can see the tip is not touching
What does it even mean to draw that?
23:41
Nonetheless it is true that $\sigma - \sigma \circ \Psi$ is a boundary chain where $\Psi : \Delta^k \to \Delta^k$ is an orientation-preserving homeomorphism; this is a good exercise!
@TedShifrin let me give you a better picture
@Lelouch Not too bad, except I woke up in the middle of the night abruptly lol
Poor a Balarka.
@BalarkaSen BTW, you're from ISI right ? I'm from CMI, and you're quite famous here :P
23:43
Oh oh, my a @Balarka is famous?
lol i'll come back later when i've got the picture drawn
@Lelouch Actually, do I know you by any chance? (Have we met in some seminar or something?)
@BalarkaSen I'm a first year moving to second year, so no. But I know people who knows people (Mohan, Akash) who did
I'm interested in differential geometric side of things, and it's pretty bad to do stuff here
like the faculty is obssessed with algebraic geometry
23:46
Crazy when an Indian has a French name to throw me off !!
*faculty as in those who do some active research
Well, you can do complex algebraic geometry with lots of differential geometric techniques.
Haha yeah, but aren't Dishant and Sushmita faculty in CMI?
They are contact/symplectic topologists if that interests you
India has an obsession with algebraic geometry in general; lots of them in ISI as well.
@BalarkaSen Huh, I don't recall seeing Dishant Pancholi in CMI. I know about Ramadas/Priyavrat
23:49
Ramadas is more into complex diff geo from what I know
and Priyavarat recently is into Data Science/CS related topology stuff
Yeah I saw they were offering a course on persistent homology haha
@BalarkaSen who ?
how's the diff geo faculty in ISI ?
Complex diff geo is cool ... clears throat
Priyavrat Deshpande and Sourish Das, @Lelouch
At least if I am remembering this right
I feel IISc has the best diff geo faculty among the UG (Vamsi, Ved, Varali, Seshadri)
23:51
Hi @Paul!!!
Hey, been a long while
Yeah IISc has a great number of geo/top people
@BalarkaSen Well they did take a course on Topological Data Analysis last year, is that waht you're referring by persistent homology ?
heya @Paul
23:53
@AlessandroCodenotti Look at the figure eight point? There's only one vector which sticks out of it, so how can there be four strands of the flowline leaving in/out of that point?
@Balarka: Hold on. One vector which sticks out? What you mean?
Doing cool things Balarka?
the flowlines of a group action?
Didn't his thing come from flow of a vector field
23:55
No.
@BalarkaSen Is it a good idea to learn about PDE/ODE if I'm interested in learning diff geo ? I only know the equivalent of Lee, Intro to smooth manfiolds
just a continuous group action
@PaulPlummer Eh, kinda.
@Lelouch Some understanding of ODEs will be necessary, yeah.
And PDE for modern-day geometric analysis.
@BalarkaSen How much of ODE is required that what's covered in a standard course (say, Arnold) ?
I skipped the section on integral flows in Lee because I lacked the prereqs
23:57
I mean for starters you just need to know like Picard-Lindelof, flows, etc. Smale-Hirsch is a good book.
you just need to know Picard-Lindelöf for that, essentially
More serious dynamics can be skipped if your purpose is to go into diffgeo.
Depending on the meaning of diff geo.
feels like he's talking to the wall here
I was thinking do Carmo, Riemannian geometry
@TedShifrin Like say I want to understand what's done in Lee, Riemannian manifolds (presumably that's a good introduction to diff geo ?)
23:59
there's a short treatment of the ODE theory you need for the basics of differentiable manifolds in the appendix of Conlon's book
@BalarkaSen I don't think you even need any of this diffeo stuff
I didn't like do Carmo because of the very little examples provided compared to Lee
So we're talking just introductory textbooks, not research. Dynamical systems is part of the story, too.

« first day (3651 days earlier)      last day (1666 days later) »