« first day (3567 days earlier)      last day (1458 days later) » 

12:01 AM
Exterior Differential Systems by Bryant, Chern, Gardner, Goldschmidt, Griffiths
 
Pig
thanks
 
Hi! Very sorry to bother. I'm trying to design a type of electronic filter (a stepped-impedance microstrip filter, not that it matters), and (being a dimwit non-mathematician) I can't figure out how to apply a polynom. technique
 
You can ask. You're not guaranteed to get an answer, depending on the question :)
 
Hollllo
 
12:14 AM
hmm I should re-type that in latex
 
What is your specific question? I am not going to try to make sense of all that.
 
sorry, yeah
 
We got programmer, Engineer or Computer scientist here?
 
idiot, primarily
I'm trying to put this into software, so programmer I suppose
 
I have seen your name in pointer.
 
12:17 AM
y'know what, screw this, I'll try to find a better derivation
sorry for all the trouble
 
Well even I am not gonna make sense out of that document lol.
 
Good luck!
 
@StupidKid wanna see real agony?
same author, 8 years later, same work
I'm convinced he let his dog typeset
 
That's unfair to dogs.
 
He defines Z and Z0 equal "system impedance" above
 
12:20 AM
😂😂😂
 
but then what the HELL does Z0^2/Z mean
 
Who knows? Variables are not defined.
Is it from electronics?
 
yeah.
anyhow bye sorry again
 
Lol.
I think I have lot to learn.......
 
no wait, I'm back with more travesty
that's it. that's the equation.
 
12:27 AM
Looks like that Equation is not fully rendered.
 
it's an image in a Word document.
 
You could even say it's half-baked.
 
That's why it is called travesty.
 
oh, and then this smartass Lee et al tries to come up with a better solution "oh, this Kapilevich dude needs a numerical optimization step, I'm going to come up with an amazing analytical approach"
but then, page 5 of 8, drops in a
"parameters were extracted from simulation for accurate design"
oh I'm sorry, that doesn't sound very analytical
laiksdfjla dmsf a
okay bye sorry again
 
Sounds, you might say, empirical.
I get it that bad writing — in whatever field — is so frustrating, and one wonders how referees accepted so many papers.
 
12:32 AM
A really quick question
Is the surface of a cube a differentiable manifold given two charts covering the north/south part?
But should it not be differentiable because it has edges intuitively?
Am I interpreting the definition wrong at some point or...?
 
It's a manifold with corners, not even a manifold with boundary.
No, you're not wrong. But Stokes's Theorem and such apply fine to such creatures.
 
Good night.
 
Nice thank u seems I might need to update my intuition a little bit.
 
Why do you need to update your intuition? You were right :P
Night, @StupidKid.
Having two Stupids that get pinged is really annoying.
 
How do you cover a cube with two charts? I don't see it.
 
12:43 AM
No, you definitely don't.
You're going to have to use individual faces, each of which is a manifold with corners.
 
Ok, that's what I thought
 
1:18 AM
@TedShifrin I couldn't sleep. A First Course in Probability (9th Edition) 9th Edition .I was thinking to buy this book but this so poor review do you have any recommendations on undergraduate probability book. I think I can't sleep thinking about it.
 
Random projection question. What does a ball in banach space projected to R^2 look like?
I suspect it will not look circular due to the norm being something more complicated than x^+y^ in general
 
@TedShifrin hey ted!
 
1:46 AM
@StupidKid: Just the title doesn't tell me. If it's Ross, I taught out of that. It's OK, but not fabulous. But good exercises.
Hi @Stan
 
Hi @TedShifrin
 
Hi NoName
 
@StupidKid we all do.
 
Hi, robjohn
 
@TedShifrin what's up?
 
1:55 AM
 
@BalarkaSen This sounded made up - I looked it up to see it's real.
 
@TedShifrin I think we all are ;-)
 
@TedShifrin I accidentally ended up in a real math class :') and i'm getting demolished
 
Oh, oh, @Stan. I thought I was the only one who demolishes you.
 
or at least studying an insane amount
Yeah but see I knew linear algebra was super important so I'm willing to get demolished for that
Whereas maximal cliques....what will I use that for?
 
1:58 AM
I don't even know what those are. And why are you taking it?
 
@StanShunpike wha' you do tha' for?
 
graph theory?
 
A course called Probabilistic Graphical Models
I thought we'd be doing applications of them. But instead we are doing these abstract properties :')
 
@robjohn Cute. Cute.
Well, perhaps if I knew the definition of a clique it'd seem reasonable. But I know not of this material.
 
probabilistic graph theory is a pretty cool topic
one of my friends wrote a thesis on it
 
1:59 AM
Does the course syllabus (not the catalog description) make it clear that he's doing applications?
 
a clique is a set of vertices such that the induced subgraph is complete
 
Nope lol. But my thesis advisor teaches it and he's an awesome guy and he thought I would get something out of it
 
Oh, that's natural enough.
 
So I took him at his word
 
Well, you may well yet.
If it's graph theory with proofs, there will be lots of induction arguments. That's not bad.
So the largest embedded complete subgraph possible.
Not sure why I put embedded.
 
2:02 AM
and in my spare time i've been studying odes for my thesis. i finally settled on using this thing called the Lotka-Volterra equations. They're pretty nifty :) I didn't realize how amazing ODEs were. It's incredible the way they can capture disease spreading
Never seen anything quite like it
 
Yes, dynamical systems is beautiful. If I didn't already tell you, read Hirsch/Smale.
 
You did! At the moment, I'm specializing in studying a particular type of differential equations called competitive Lotka-Volterra equations
@TedShifrin are you familiar with those?
 
I have seen them, but "familiar" is a stretch.
 
2:42 AM
Hello, this is more of a stackexchange specific question but is there any way to see all the posts from all my 'watched' tags?
 
 
4 hours later…
6:30 AM
The argument of $z$ is $\theta$ and the modulus of $z$ is 1. Find the $arg\left( \frac{1+z}{1+\overline{z}}\right)$
I tried so hard, but couldn't anything
I tried everything $z=. e^{I\theta},~~ z= \cos\theta + i \sin\theta, ~~z= x+iy$
but none of them helped.
Hello Ted
 
Did you play around with the fraction?
How do you usually divide complex numbers?
 
Yep
 
or better still, draw a picture
there's a really neat geometrical solution
 
We try to eliminate $i$ from denominator
Oho Leaky! draw ! Please
is it neat or beautiful ?
 
are you telling me to draw?
 
6:35 AM
Much as I love pictures, a technique that would always work might be good.
 
no, I'm asking you to guide me to draw
 
But I agree that a picture is great here (but you have to do arithmetic with angles).
 
Ted sir please tell me algebraic method also
 
So how do you actually compute $z/w$?
In the $x+iy$ form, I mean?
 
I multiply by the conjugate of $w$ to both $z$ and $w$
 
6:38 AM
OK, so did you do that in this case?
 
YEP
 
And what did you have?
 
just a second
$$\frac{1+x+iy}{1+x-iy} \times \frac{1+x+iy}{1+x+iy} = \frac{(1+x)^2 -y^2 +2iy(1+x)}{(1+x)^2 +y^2}$$
 
I would have just used $z$ in there.
 
okay. but what is conjugate of $1+\bar z$ ?
 
6:43 AM
You tell me.
 
Leaky couldn't resist.
 
@TedShifrin is it $\overline{1+z}$ ?
 
No.
 
Well, $1+\bar z = 1 + (x-iy)$
 
6:45 AM
Stop that.
 
that is $(1+x) -iy$
LOL
 
What is the conjugate of a sum?
 
you're a very strict teacher
@TedShifrin sum of conjuagtes
 
So what is the sum of the conjugates?
In this case, it really is better just to subtract the arguments of the original complex numbers and use symmetry ... without being as clever as Leaky was. But you can still hack out the algebra.
 
OH! yes, I got you $$ \overline{1+\bar z } = 1 +z$$
 
6:48 AM
Right.
 
now I'm sure Ted will be more than happy to explain the geometrical solution
 
If you're going to go this route, you should definitely put in $z=e^{i\theta}$.
 
So, I will multiply by $(1+z)$ ha
?
 
Use $z=e^{i\theta}$ and then think about what argument is. And you'll need double angle formulas. I think Leaky wins hands down that you should use some geometry to start with.
 
Please explain that geometrical solution.
 
6:53 AM
you don't need any of those
 
But it's reassuring that it works out using the algebra/trig.
 
you just need the fact that the angle at circumference is half the angle at centre
 
You don't need to be as clever as Leaky.
 
I'm Leaky
 
Draw a point on the unit circle and go one unit to the right. What is the angle from the origin to that point?
 
6:54 AM
I'm tight
 
(that's wh-)
 
Leaky's picture is telling you to think about $z+1$ as $z-(-1)$.
But you don't need to be that clever.
 
"go one unit to the right" do you mean horizontally ?
 
Yes. That's what $+1$ means.
 
But then I won't be on circle anymore
 
6:56 AM
So what?
We're plotting the point $1+z$.
 
Yes, I did that
 
And if $z=e^{i\theta}$, I'm asking what the argument of $1+z$ is from the picture.
I leave you to it. Good night, all.
 
Good Night sir
 
7:35 AM
WTH is integral index lmao
In past I only look alternative definition but it doesn't tells me what is integral index
 
@Ted @Pig Let me do a coordinate calculation. I will identify $T\Bbb R^3 = \Bbb R^3 \times T_0 \Bbb R^3$, and $TT\Bbb R^3 = T\Bbb R^3 \times TT_0\Bbb R^3 = \Bbb R^3 \times T_0 \Bbb R^3 \times T_0 \Bbb R^3 \times T_0T_0\Bbb R^3$ and thus $TTT\Bbb R^3$ with $$\Bbb R^3 \times T_0\Bbb R^3 \times T_0\Bbb R^3 \times T_0T_0\Bbb R^3 \times T_0 \Bbb R^3 \times T_0T_0\Bbb R^3 \times T_0T_0\Bbb R^3 \times T_0T_0T_0\Bbb R^3$$
 
Is index power?
Is index and indices the same?
 
Given a map $f : \Bbb R^3 \to \Bbb R^3$, the induced map $\mathbf{D} f : T\Bbb R^3 \to T\Bbb R^3$ is $\mathbf{D} f (x, u) = (f(x), Df_x(u))$. The induced map $\mathbf{D}^2 f : T^2 \Bbb R^3 \to T^2 \Bbb R^3$ is $\mathbf{D}^2 f(x, u_1, u_2, v) = (f(x), Df_x(u_1), Df_x(u_2), Df_x(v, u_2))$. Let me try the third iteration:
Sorry, I meant $D^2 f_x(v, u_2)$ in the last coordinate in the last line.
This is the Hessian, $v^T Hf^i_x u_2$, on each coordinate $f^1, f^2, f^3$ of $f$
So the third iteration is $\mathbf{D}^3 f : T^3 \Bbb R^3 \to T^3 \Bbb R^3$, let's denote an element of $T^3 \Bbb R^3$ lying over the point $x \in \Bbb R^3$ as $(x, u_1, u_2, v_1, u_3, v_2, v_3, w)$ where $u$'s are tangent vectors, $v$'s are double tangent vectors aka tangent vectors on $T_x \Bbb R^3$ i.e., in $T_u T_x\Bbb R^3$ aka variations, and $w$'s are triples tangent vectors, aka tangent vectors on $T_u T_x \Bbb R^3$, i.e., in $T_v T_u T_x \Bbb R^3$
$v$'s are variations of $u$'s and $w$'s are variations of $v$'s which are variations of $u$'s. Note the difference between on and in I have emphasized for clarity
Explicitly, this seems to be $\mathbf{D}^3 f : T^3 \Bbb R^3 \to T^3 \Bbb R^3$, given by $$\mathbf{D}^3 f(x, u_1, u_2, v_1, u_3, v_2, v_3, w) \\ = (f(x), Df_x(u_1), Df_x(u_2), D^2 f_x(v_1, u_2), Df_x(u_3), D^2 f_x(v_2, u_3), D^2 f_x(v_3, u_3), D^3 f_x(w, v_3, u_3))$$
 
7:58 AM
Sht wtf I asked lmao
😂😂😂
pure idiot
 
To be extremely explicit what I mean by derivatives again, $Df : \Bbb R^3 \to \text{Hom}(\Bbb R^3, \Bbb R^3)$ is the pointwise linear operator, $x \mapsto (u \mapsto Df_x(u))$, $D^2 f : \Bbb R^3 \to \text{Hom}(\Bbb R^3, \text{Hom}(\Bbb R^3, \Bbb R^3))$ is $x \mapsto (u \mapsto (v \mapsto D^2 f_x(v, u)))$ is the derivative of this operator.
So $D^3 f : \Bbb R^3 \to \text{Hom}(\Bbb R^3, \text{Hom}(\Bbb R^3, \text{Hom}(\Bbb R^3, \Bbb R^3)))$ is $x \mapsto (u \mapsto (v \mapsto (w \mapsto D^3 f_x(w, v, u)) ) )$ is the further derivative of $D^2 f$
So $\mathbf{D}^3 f$, the "$3$-tangent prolongation", seems to give me exactly the same information as the "$3$-jet prolongation" of $f$, namely, the value of $f$ at $x$ and the first three derivatives.
I don't see what goes wrong.
It's linear in all the variable $u_1, u_2, u_3, v_1, v_2, v_3, w$ so that gives me $9$ degrees of freedom each! So the fiber over the point $x$ is $63$ dimensional.
This is huuuge so I don't get the dimension issues you all were having
I am of course repeating lots of things, that's why my object is bigger
Can you tell me if I am insane and doing something wrong? I sure as hell don't see what
 
8:52 AM
Why do we even need to differentiate 1/(1-x) m times in order to established the binomial theorem for negative integral index. It's superfluous.
It doesn't give me any information.
Anyway I did and can't see anything. It's $(-1)^2m(1-x)^{-1-m}m!$.
I'd rather start from the -1 and start to perform induction on it.
Guys anyone see any useful information?
if I see differentiation m times what I see lmao TBH.
I think book is just joking with me.
It is trollin me.
 
9:22 AM
y0
 
lol edward scissorhands
 
9:42 AM
Hi @Edward
 
10:18 AM
Hiya @Alessandro
 
Hello Ed
@robjohn Sir you there?
 
@Knight what's up?
 
@robjohn @Hello
Sir I need your help in a problem, if you're not asleep.
 
10:34 AM
@StupidKid Apply Taylor's Theorem?
 
4 hours ago, by Knight
The argument of $z$ is $\theta$ and the modulus of $z$ is 1. Find the $arg\left( \frac{1+z}{1+\overline{z}}\right)$
 
@Knight Note that when $|z|=1$, $\bar z=\frac1z$
 
YES
 
so $\frac{1+z}{1+\bar z}=z$
 
because we have $z=e^{i\theta}$ and conjugate $\bar z = e^{-i \theta }$
@robjohn How?
 
10:40 AM
@Knight plug in $\bar z=\frac1z$
 
$$ \frac{ 1+ z } { 1+ \frac{1}{z} } $$
 
and simplify
 
Oh Yeah!
WOW! sir you're great
sir can we do it in any other way?
like explicitly putting $z= e^{i \theta}$
 
@Knight you can, but it will be cancelling the factor of $e^{-i\theta}$ from the denominator, same as canelling the $\frac1z$.
 
$$1+z = 1+ e^{i \theta} \\ 1+\bar z = 1+ e^{-i \theta }$$
$$\frac{1+e^{i\theta}}{1+e^{-i \theta}} \\ \frac{ (1+e^{i\theta})^2}{1+e^{-i\theta} + e^{i\theta} + 1}$$
What should I do now?
I tried multiplying by the conjugate of $1+e^{-i \theta}$ but it didn't remove $i$ from the denominator
$$ e^{i\theta} + e^{-i\theta} = e^{i\theta} + \frac{1}{e^{i\theta}} \\ = \frac{e^{2i \theta} +1}{e^{i\theta}}$$
@robjohn Sir please help
 
11:02 AM
@Knight Multiply numerator and denominator by $e^{i\theta}$
 
WHY?
it's not a conjugate
 
I'm not talking about your last equation
I replied to the one I was referring to
 
Oh! but Astyx why I didn't get a pure real number when I multiplied it by the proper conjugate
?
 
@Knight $\frac{1+e^{i\theta}}{1+e^{-i\theta}}=e^{i\theta}\,\frac{1+e^{i\theta}}{e^{i\theta}+1}=e^{i\theta}$
 
Oho! Yes.
:-)
 
11:04 AM
$e^{i\theta} + e^{-i\theta}$ is a real
It's $2\cos \theta$
 
But why not in Euler form?
how to write $2\cos \theta$ in Euler form?
 
$e^{i\theta} + e^{-i\theta}$
 
yes
So, I have $$ \frac{ (1+e^{i \theta})^2}{2+ e^{i\theta} + e^{-i\theta}}$$ see we can't go beyond this step if we were to follow only the standard steps (standard means "taught" steps)
 
11:50 AM
@robjohn why do I need to apply taylor theorem?
 
@Pig @Ted What you said got me thinking about what the difference between $J^2(M)$ and $TTM$ (though please note I wasn't arguing they are the same in the first place): I believe they are dual in the following sense. $TTM$ is the space of "2-flares", germ of a variation of germ of a curve $\Bbb R \to M$, whereas $J^2(M)$ is space of 2-jets, germ of a variation of germ of a map $M \to \Bbb R$
 
I think by induction I can prove it for negative index.
 
Which is an amusing thing to say I suppose
And really when I say germ I mean a 1-truncated germ
 
the last part of that last sentence is an absolute abomination
 
I already did it btw B=)
 
11:54 AM
@Thorgott can you actually imagine a germ as anything other than a variation, which is inherently inaccurate as it just keeps track of the 1st order part of the germ? :P
 
Now I am proving it for real index and then blancmange function index. Then I can say bye to real analysis course. So excited about measure theory and Group theory right now.
 
I would imagine the k-truncation of the germ as a k-fold variation, variation of a variation of a ...
 
At least I can discuss higher lv math. =~]
Then topology
 
I don't even know what a variation is
 
Take a curve $\gamma : \Bbb R \to M$ with $\gamma(0) = p$. A variation of $\gamma$ is a vector field $X$ along the curve such that $X(p) = 0$
 
11:57 AM
Hi who can help with these
 
This is because you can variate the curve in the direction of $X$, namely, define $\gamma_t : \Bbb R \to M$ for $-\varepsilon < t < \varepsilon$ by $\gamma_0= \gamma$ and $d\gamma_t(s)/dt|_{t = 0} = X(\gamma(s))$
these variations are not unique but all such variations are "tangent at $\gamma_0 = \gamma$" because of the condition
one could imagine a big space of all curves passing through $p$ in $M$, where $\gamma$ is a point and $X$ is to be thought as a tangent vector to the point $\gamma$ in that big space
 
you're defining the curves by how they vary with the parameter?
ouch my brain
 
haha im defining a family of curves such that the original gamma is the "center curve" of the family, and if you move away from gamma to another nearby curve in the family you're moving in the direction of X
my X is their J
 
12:12 PM
What map is an isomorphism between the two fields
$\mathbb{F}_3[x](x^2+1)$
and $\mathbb{F}_3[x](x^2+2x+2)$
?
 
ok, that makes sense geometrically
what exactly is $d\gamma_t(s)/dt\vert_{t=0}$ here?
 
$\gamma_{-}(s) : (-\varepsilon, \varepsilon) \to M$ is a curve passing through $\gamma_0(s) = \gamma(s)$
take the derivative at $0$
these are the "transversal curves" to the variation
 
Oh no are you still doing that
 
Doing what bro
 
jet nonsense
 
12:19 PM
More or less
 
That gives me a linear map $T_0(-\varepsilon,\varepsilon)\rightarrow T_{\gamma(s)}M$, but $X(\gamma(s))$ is an element of $T_{\gamma(s)}M$. Do I just identify by looking at the image of the standard basis?
 
Wait they taught you the derivations definition of tangent vectors but not curves definition
Weird, but yeah, that's what you do
Look at image of the coordinate vector $\partial/\partial t \in T_0(-\varepsilon, \varepsilon)$ under $d\gamma_{-}(s)$
 
Hello
 
ohh, that's how you mean it
I know the curves definition
 
OK whew
 
12:23 PM
just getting lost amidst all these identifications
 
Why is (closed) Poincaré dual of $S^1$ in $\Bbb R^2 - 0$, $0$?
 
What is your definition of Poincare dual
 
Let $S$ be a closed k-submanifold of $M$. The closed Poincare dual of $S$, $\eta_S$ is given by $\int_S i^*\omega = \int_M \omega \wedge \eta_S$ for all $\omega \in H^k_c(M)$.
 
If it's the de Rham definition, the Poincare dual will be a closed 1-form which integrates to zero along $S^1$ so will be exact
 
Basically, $\omega \mapsto \int_S i^*\omega$ defines a linear functional on $H^k_c(M)$, so by Poincare duality, this gives you an (n-k) form, that form is Poincare dual.
@BalarkaSen which integrates to 0? Never seen that definition.
 
12:32 PM
This is not definition, this is a calculation. Say $\eta$ is PD[S^1], then $\int_{S^1} \iota^* \eta = \int_{\Bbb R^2 - 0} \eta \wedge \eta = 0$.
But a closed 1-form on $\Bbb R^2 - 0$ which integrates to $0$ along $S^1$ is exact
So PD is the 0 cohomology class
 
I didn't know Joseph Silverman was active on MSE. Nice. (Just noticed that he answered a question looking for an example of something).
 
@BalarkaSen Thanks. This is what I needed.
 
What map is an isomorphism between the two fields
$\mathbb{F}_3[x](x^2+1)$
and $\mathbb{F}_3[x](x^2+2x+2)$
?
Is it the identity?
 
that notation doesn't make sense to me unless you put a / somewhere
 
What map is an isomorphism between the two fields
$\mathbb{F}_3[x]/(x^2+1)$
and $\mathbb{F}_3[x]/(x^2+2x+2)$
?
fixed it
 
12:43 PM
@BalarkaSen Does that mean for all n-manifold with n even, the Poincare dual of n/2 submanifold is 0? Because I can take $\omega = \eta$?
 
You needed the structure of $\Bbb R^2 - 0$ crucially; it needn't be true that a closed form which integrates to zero along a submanifold is exact
Also you need n/2 odd for my argument
 
oh right.
 
Think of $T^2$ with $S$ being the meridian. A basis for $H^1(T^2; \Bbb R)$ is given by $dx$, $dy$ and $PD[S] = dy$
$dy$ integrates to zero along $S$, which is the $x$-curve
Certainly not a exact form
 
well, they're algebras over $\mathbb{F}_3$ of dimension $2$, so you want a degree $1$ polynomial $f$ such that $f^2+2f+2+(x^2+1)=0$ (or vice versa)
 
@feynhat Detour: Have you heard of Borel-Moore homology
 
12:50 PM
Nope.
I was reading Nicolaescu's notes on Poincare duality, he refers to elements of $(H^k_c(M))^*$ as Borel-Moore cycles.
 
Right. PD for compactly supported cohomology is $H_n(X) \cong H^{n-i}_c(X)$; the Borel-Moore theory gives a PD of the form $H^i(X) \cong H_{n-i}^{BM}(X)$
I can tell you the definition if you want. I have never worked through it very precisely
 
I'll stick with de Rham for now.
 
chicken :P
fine
TBH I can never think of (co)homology in terms of forms, it's always singular chains and singular cochains for me
its a pity because that POV is supposed to be powerful
 
@BalarkaSen How did you get the generators for $H^1(T^2)$? Is there a better way than chasing forms on Mayer-Vietoris?
(or maybe you translated the singular generators to forms using de Rham isomorphism :P)
 
I just pushed forwards the $\Bbb Z^2$-equivariant forms $dx, dy$ on $\Bbb R^2$ to $T^2$ by $\Bbb R^2 \to \Bbb R^2/\Bbb Z^2$, and I know $H^1(T^2) \cong H_1(T^2)^*$ is generated by the meridian and longitude cycles
yeah :P
 
12:58 PM
lol
 
I translated via de Rham isomorphism and dualized by universal coefficient theorem to be precise
Didn't do PD during translation procedure
just learn it in terms of chains man do borel moore with me
what do i need to convince you this supreme point of view
dont let ted hear me say that tho
 
omg you're a fanatic
 
ill brand you as a dirty analyst if you dont learn borel moore
i swear to god
 
@BalarkaSen I will... and from you.
but not now.
 
haha ok i was just ribbing
i think the point is geometrically one should expect the PD of $[S^1]$ in $\Bbb R^2 - 0$ to be the infinite ray, say $\theta = 0$
But that's a nonsense class in standard homology theory
Borel-Moore allows possibly infinite chains so it happens to become a class
 
1:07 PM
@BalarkaSen Why would you expect that?
is it because PD of ray is circle?
 
in my imagination, if you have two submanifolds $N_1, N_2$ of $M$ which have complementary dimension ie $\dim N_1 = \mathrm{codim}\, N_2$, $N_1$ and $N_2$ intersect transversely at a point, then they are PD because then under the intersection pairing $H_*(M) \otimes H_{n - *}(M) \to \Bbb R$, $[N_1] \pitchfork [N_2] = 1$
so in the isomorphism induced by this nondegenerate form $H_\ast(M) \to H_{n-*}(M)^*$, $[N_1]$ gets sent to $[N_2]^*$
@feynhat ah yeah that's true. actually that is what Borel-Moore theory gives, ray is an actual homology class
but whatever this is too much gibberish
 
@BalarkaSen This is the idea that I need to get used to.
 
there should be a way to interpret these kind of compactly-supported dualities as homology-cohomology duality for the one-point compactification $M^+$ where $M$ is a noncompact manifold
even though $M^+$ is not actually a manifold, PD makes sense
I think this is a major motivation for a lot of work in the late 70s by Goresky-MacPherson where they introduced intersection homology, to define PD for sufficiently well behaved singular spaces
this is pie in the sky, i will try to work out an example and tell you about it afterwards, nevermind my bullshitting
 
1:25 PM
@Balarka suggest me some music
 
trv kvlt?
do you know Eloy
 
Only by name
 
Dude listen to Ocean
 
Classic 70s prog, noice! Any album in particular?
 
Great album
 
1:27 PM
Thanks, I'm going to that
 
yeah A+ prog rock
Cheers
 
2:03 PM
who would have thunk that you can find a concrete proof in nLab
this proof avoids complex analysis by making use of the fact that we already have a taylor expansion
and the open mapping theorem only depends on the existence of taylor series
it is even more convenient here, since the taylor expansion is finite
 
I don't see how this essentially relies on the open mapping theorem
I think this is the same idea as the Fefferman proof
 

« first day (3567 days earlier)      last day (1458 days later) »