« first day (2675 days earlier)      last day (2345 days later) » 

2:00 AM
@orbit-stabilizer it didn't say that
 
That's not even relephant.
 
"Let f be a continuous function that maps the unit interval [0, 1] in R to itself."
 
oh, right
@TedShifrin how would you prove that symmetric (alt. self-adjoint) matrices are diagonalizable?
 
Um, that's a nontrivial theorem. My favorite proof uses Lagrange multipliers.
 
I thought induction would work
 
But you can do it by complexifying and using a hermitian inner product.
You need induction, but you need more than that.
 
Maybe I am missing something in Arnold and Hamiltonian vector fields ARE sections of the normal bundle to xi.
@MikeMiller are you familiar with the name "Hamiltonian" vector fields.
 
@TedShifrin orthogonal complement?
 
Leaky, how do you know that the eigenvalues are all real?
 
by pretending that the matrix is complex
 
2:03 AM
@anakhronizein I am but only in the context of symplectic geometry.
 
oh
well, then, sure.
I just said that up there.
 
right
 
I still prefer Lagrange multipliers.
 
how does that work?
 
Maximize $f(x)=Ax\cdot x$ subject to the constraint $\|x\|^2 =1$.
 
2:05 AM
Lagrange multipliers are always fun
 
"Symplecetification of vector fields is an isomorphic map of the Lie algebra of contact vector fields onto the Lie algebra of all locally hamiltonian vector fields with hamiltonians which are homogeneous of degree 1"
- Arnold
 
They're among my favorites ... especially because I gave a garbage proof on an exam my freshman year. I never did that again :P
 
That sounds approximately like what we are looking for.
 
@TedShifrin and which criterion of diagonalizability are we using?
 
And that certainly sounds like Arnold
 
2:05 AM
Basis of eigenvectors.
 
@KevinDriscoll reacts
 
So now to find out what the symplectification of vector fields is.
 
@Semiclassic: You mean a sentence where you drop dead when you can't take a breath?
 
pretty much
 
@orbit-stabilizer We also know $f'$ is continuous
 
2:06 AM
Heeeey!
 
Oh, $f$ is uniformly continuous
 
Intuitively, how would you explain what an aperiodic matrix is?
 
Well, that's true, but still not the direct observation I wanted, @orbit.
 
@nbro Herou
Hello
 
oh. a follow up question from earlier
like I said earlier, I can write $f(x,y,z)=\det(Ax+By+Cz+D)$ with real symmetric $D$.
 
2:08 AM
Yes, $f'$ is continuous..
 
@TedShifrin someone had me prove that I can have $y=Px$ such that $\langle Ax,x \rangle = \|y\|^2$ and $\langle Bx, x\rangle = \langle y,Dy \rangle$ where $D$ is diagonal. that's quite cool.
 
And thus you can apply theorems about continuous functions to it @orbit-stabilizer
 
of course, $A$ and $B$ are symmetric, with $A$ positive-definite
 
@orbit: Right, so $f'$ has a maximum on $[0,1]$.
 
is it obvious that $f(x,y,z)\geq 0$ is a convex set?
 
2:08 AM
Ah, it is bounded
 
No, @orbit ... you're still missing the big theorem.
 
Extreme Value Theorem?
 
I don't know what you're talking about, Leaky.
 
blah, it can't be obvious since I can write $-f(x,y,z)\geq 0$ like that as well.
 
You certainly need more hypotheses, Semiclassic.
 
2:09 AM
right.
 
@TedShifrin given $A$ positive-definite symmetric, $B$ symmetric, prove that there is an invertible $P$ such that $\langle Ax, x \rangle = \|Px\|^2$ and $\langle Bx, x \rangle = \langle Px,DPx \rangle$ for all $x$, where $D$ is a diagonal matrix
 
@orbit-stabilizer That says it's bounded and attains its maximum
and minimum
 
I don't want to think about it, Leaky. I'm about to leave again.
 
@orbit-stabilizer needs to do more topology :P
 
Please, tag me if you're writing me regarding my question.
 
2:11 AM
@TedShifrin I proved it, I just want to say that it's a cool theorem
 
I don't see why it's cool.
It looks technical.
 
Since I can't follow now your discussions...
 
Right. So the derivative attains its maximum value on the interval. Okay. And this maximum is less than 1.
 
@nbro: I don't even know a definition.
 
@TedShifrin you're dealing with two symmetric matrices together
 
2:11 AM
The absolute value of it anyways
 
two symmetric matrices that can be made "nice" by the same $P$
 
So you can diagonalize one with respect to the other, since one is positive definite. But why is this cool?
 
@TedShifrin because I didn't know it
 
Oh.
 
@TedShifrin I can tell you it is related to Markov chains and the PageRank algorithm...
 
2:13 AM
That's pretty vague.
 
hrm. I had hoped that I'd be able to say $A,B,C,D$ were positive semidefinite
but it doesn't look like that's true here
 
@Semiclassical what's the question?
 
Scroll up, Leaky.
It's the responsibility of those of us who interrupt to scroll up.
 
@orbit-stabilizer Yarp
 
lol
it's a while up, to be fair
 
2:16 AM
@TedShifrin What is this a response to?
(/s)
 
4 hours ago, by Semiclassical
$$f(x,y,z)=4xyz-(x+y+z-1)^2= \frac{1}{4}\begin{vmatrix} 1 & 2x-1 & 2y-1 \\ 2x-1 & 1 & 2z-1 \\ 2y-1 & 2z-1 & 1\end{vmatrix}$$
 
Leaky's "what's the question?" question.
 
(Kidding)
 
@AkivaWeinberger about what?
 
@Semiclassical Oh, no, Semiclassic, I was referring to that.
 
2:17 AM
I want to show that $K=\{(x,y,z)\in[0,1]^3|f(x,y,z)\geq 0\}$ is convex.
 
DogAteMy: You had better keep me posted on your collegiate progress.
Or else.
 
@AkivaWeinberger I want to show $|\frac{f(x)-f(y)}{x-y}| \leq M$. That looks very close to being the derivative... if I fix x and let y -> x. The derivatives are all bounded. But I don't know how that helps me.
 
Not much has happened between sending in my Yale app and now
 
@orbit. You already said you needed the Mean Value Theorem. What did you tell us about $f'$?
 
Did I tell you sent in my Yale app?
 
2:19 AM
Yes, DogAteMy, I knew that much ... Did you email my former student/friend?
 
AH CRAP i forgot
 
growls louder
2
 
@TedShifrin is there a polynomial map $\Bbb C^n \to \Bbb C^n$ which is easy to check injective but not surjective?
 
Not that it matters. But I thought you should get a math/music connection.
 
Ted intensifies
6
 
2:19 AM
(thus making Ax-Grothendieck useful?)
 
He graduates soon.
 
It's continuous.... and it achieves a maximum.
 
So what do you know about $f'(c)$?
 
$f'(c) < f'(a)$, where $f'(a)$ is the maximum
less than or equal to
 
@TedShifrin what if the derivative isn't continuous?
 
2:21 AM
Good. And what do you know about $f'(a)$?
It was, Leaky.
 
"what if"
 
$|f'(a)| < 1$
 
24 mins ago, by orbit-stabilizer
Let f be a continuous function that maps the unit interval [0, 1] in R to itself. Assume
that f has a derivative f' which is defined and continuous on [0, 1] and that |f'(x)| < 1
for x ∈ [0, 1].

Show that there is a constant M < 1 such that for all x, y in [0, 1],
|f(x) − f(y)| ≤ M|x − y|
For reference
 
save that for after the question at hand is shown
 
2:21 AM
@Semiclassical it's already solved if you ask me
it's just one step
 
@LeakyNun I'm willing to bet bounded derivatives attain their maximums
 
We didn't ask you.
tired of condescending attitudes by people who think they are superior
DogAteMy: So derivatives cannot have jump discontinuities. Does it follow?
I don't believe a bounded function on a closed interval must attain its max. But if you know it's a derivative, must it?
 
@TedShifrin I hope so
Hm
 
In R yes
it must attain its max
 
How so?
 
2:24 AM
Derivatives need not be continuous @orbit-stabilizer
I don't know if you've learned that yet
If not, it's quite shocking, and I hope you were sitting down
 
Why should you expect derivatives to be continuous?
Just because the speed of a car is apparently continuous? :P
 
Because the only time they're not is when stuff is really weird
and siney
 
That's far from the only time.
 
or cosiney
 
"I don't believe a bounded function on a closed interval must attain its max."

A function on a closed and bounded interval attains its max in R.
 
2:26 AM
NOOOOO @orbit
 
cts
sorry
it has to be cts
 
cts?
 
Cts=continuous?
 
continuous
 
right
 
2:27 AM
Right, that's the point, I was asking about potentially discontinuous functions
 
but we do know that a derivative satisfies the intermediate value theorem
by some magic
 
not magic
 
Darbosomething
 
Darboux
 
Darbau?
 
2:28 AM
I don't know if orbit knows that.
 
lol
 
Darboux
 
No... We're only on chapter 5 of Rudin. covered up to differentiation
 
I like DogAteMy's rephrasing of the question earlier.
 
@orbit-stabilizer have you covered compactness?
 
2:28 AM
Durboh
 
Wait if "ou" is pronounced oo in French why is "oux" oh
 
DogAteMy: It's not.
 
Oh is it just IVT?
 
It's also oo.
 
2:29 AM
No? So is Darboux Darboo?
 
eau or eaux are oh
 
@orbit-stabilizer then you know that a continuous function maps a compact set and compact set
 
@TedShifrin Oh
 
Yes.
 
and here compact = closed and bounded
 
2:29 AM
No, oo. @Akiva :P
 
well, /o/
I meant, a compact set to a compact set
pebkac
 
And Dābu in Japanese I imagine
 
Okay, I'm just totally lost right now. I think I'll come back to that question later or ask someone. Thanks for the help!
 
I won't roll eyes because of ignorance of Japanese.
 
2:31 AM
I'm wrong, it's Darubū
 
@orbit: You had it.
 
Dammit Leaky
 
sniped :P
dem
 
I am trying to do big O notation
 
You said that $f'(c)\le f'(a)$ for all $c$. And what do we know about $f'(a)$?
 
2:32 AM
1/n + 15 - 3 is O(n^3)
 
Can you actually read Katakana? @LeakyNun
 
@ErikHumphrey $\Huge{O}$
@AkivaWeinberger yes, and hiragana
and some kanji
 
@Erik: It's $O(1)$.
 
@TedShifrin The only thing I know is that $|f'(a)| < 1$.
 
2:32 AM
@AkivaWeinberger but I don't know most kanji's meaning/pronunciation in Japanese
 
Stop sniping me Ken
 
I only know them from Chinese
 
Ah makes sense
 
what do I do with the 1/n?
 
That's bounded by 1
 
2:33 AM
Sorry, trying to prove that it's not O(n^3)
so the first step is that i'm adding n^3 to everything
 
Wait, @Erik: As $n\to\infty$?
 
??? + 15n^3 - 3n^3
 
As n goes to infinity or as n goes to 0?
 
You have a typo in the original, @Erik. Check.
 
n is all positive real numbers
 
2:34 AM
$1/n$ goes to $0$. It's of no issue as $n\to\infty$.
 
is there any connection between categorical limit and analytical limit?
 
Yes. It is if you consider n -> inf.
 
There has to be a typo in there. Why did they write 15 - 3 instead of 12?
 
Good question
 
2:35 AM
You're making this way too hard.
 
Is there any connection between an analytical limit and the actual word in English?
 
it's even O(1)
 
Assuming no typo, is $1/n + 12$ $O(1)$? Then of course it's $O(n)$ or $O(n^2)$ or $O(n^3)$.
 
I am using lecture notes from a previous year because they make the concept more easy to understand
 
I said that already, orbit :P
 
2:35 AM
I'm surprised you knew it was O(1) that quickly
 
<---- old
@orbit doesn't have the same excuse.
$O(1)$ means BOUNDED.
 
I do not know what this word means
 
Gahhh, so $|f'(a)| < 1$ - is that what you're getting at?
 
@Erik: It means that your function is always less than some positive number.
 
also, what browser extension can I install for this funky in-line LaTeX you are using?
 
2:37 AM
In your case, it would be 15, say.
 
ah
it's in the motd
 
See the link on the right above.
 
sniped :P
 
I'm leaving in a moment, so I shall stop sniping.
 
much better
 
2:37 AM
@orbit: Absitively.
That's your $M$.
 
@AkivaWeinberger well, it's the limit that you (can't) reach, for increasing sequences
 
What should I put before the $\leq$ symbol?
or shold I rework my method entirely
 
$1/n + 12 \le 15$ for all $n$.
 
https://en.wikipedia.org/wiki/Big_O_notation

I found reading the definition very helpful
 
We'd better be thinking about $n\to\infty$ and say $n\ge 1$.
 
2:39 AM
@TedShifrin yeah, I'm going to have to rework this. Because I don't understand.
Thanks though
 
> I found reading the definition helpful
words to live by
 
heya @anon. Hugs :)
 
heya
 
ehh? why ≤ 15?
 
Yes, I told one of our denizens earlier that his major change should be to master definitions and keep neat notebooks.
@Erik: 13 should be fine if we have $n\ge 1$. I was just being silly.
But of course $13\le 13n^3$ for all $n\ge 1$.
 
2:40 AM
I feel there can only be one right answer
 
NO.
 
for all n > 0
no?!
 
Yeah, but they mean integers, not real numbers.
 
you have a point
 
$n$ always denotes an integer :) Sloppy people.
OK, I'm out. Going to cook dinner. You all learn without me :P
 
2:41 AM
now I must know why it could be 13 or 15
 
Let $n \in \Bbb R$
 
It can be anything bigger than or equal to 13 ...
Figure out why.
 
the resources to figure it out are unavailable
 
Youtube!
 
the $\frac{1}{n}$ is all that's throwing me off
this is the ancient egyptian solution model I found online:
for a different question, of course
 
2:43 AM
What are you trying to show?
 
how is the evaluation homomorphism defined for $\Bbb R(X)$?
 
f(x) = O(g(x)) as x-> inf........ what does this mean @ErikHumphrey?
Can you write the defintion?
 
@LeakyNun it isn't, as the denominators can have roots
 
I am attempting to apply that to prove I am attempting to apply that to prove that $\frac{1}{n} + 15 -3 \text{is not} O(n^3)$
ah, forgot the spacetex
 
2:44 AM
technically it's not defined on all of $\Bbb R(x)$, but it's partially defined
and certainly defined on $\Bbb R[x]$
 
@anon then is it still a homomorphism?
oh, and is $\vec 0$ an eigenvector?
 
@ErikHumphrey it is $O(n^3)$!
 
it is?!
what about the part where it was constant?!
 
Unless you are talking about it $\theta(n^3)$
 
@LeakyNun it's an $\Bbb R$-algebra homomorphism if we restrict the domain to a certain subalgebra of $\Bbb R(X)$ (the localization at kernel of $\Bbb R[x]\to\Bbb R$)
 
2:46 AM
it's $O(1)$ and $O(n)$ and $O(n^2)$ and $O(n^3)$
 
also, eigenvector of what?
 
@anon of any matrix
 
@ErikHumphrey tell me what the definition is
 
it's kind of paradoxical to me
 
@LeakyNun 0 is an eigenvector iff it's singular, right?
 
2:47 AM
f(x) = O(g(x)) as x-> inf........ what does this mean @ErikHumphrey?
 
I know, I said eigenvector
 
@LeakyNun all linear algebra texts I know define eigenvectors as non-zero
 
pretend I said $\vec{0}$ instead of 0 if it suits you :P
well, then I guess they're not eigenvectors
 
@orbit-stabilizer I don't know! It's discrete math and asymptotic analysis, not calculus!
 
@MatheinBoulomenos @anon it's kind of paradoxical to me: if it isn't an eigenvector, then how do the set of eigenvectors (with the same eigenvalue) form a subspace?
 
2:47 AM
oh, right, the vector is 0, not the value
bleh
 
How can I prove it is O(n^3)?
 
I'm just derping
 
@orbit-stabilizer that means the limit of the curve is the big O notation of another function as x approaches positiive infinity
 
@LeakyNun the set of eigenvectors, together with $0$ form a subspace
 
@ErikHumphrey $C\le Cn\le Cn^2\le Cn^3$ for $n$ bigger than $1$
 
2:49 AM
@MatheinBoulomenos hmm
 
Big-O is calculus....
 
but you lose elegance
 
it's really a matter of definition
 
meh
 
You want to say "let $v$ be an eigenvector" a lot of times in proofs
this allows you to omit the non-zero
 
2:50 AM
it's still quite unclear what to do with the given 1/n
i'm not the best with fractions
 
so you get the elegance at other places
 
@ErikHumphrey $1/n\le 1$ so $12+1/n\le 12+1$ (again, when $n$ is bigger than $1$)
 
@Leaky I don't usually talk about the space of eigenvectors, I say the eigenspace
 
given the denominator i suppose this would work:
 
@Daminark what is an eigenspace
 
2:50 AM
The eigenvectors to a given eigenvalue along with 0
 
hmm
 
we say $f(n) = O(g(n))$ if $\exists n_0, M \in \mathbb{N}$ such that $|f(n)| \leq M |g(n)|$ when $n > n_0$.
does that make sense?
 
oh man i did that eigenspace stuff recently
 
when n > n_0 ofc
 
what's...what's M?
 
2:51 AM
a constant
 
Some constant, it just means the growth of one is bounded by that of another
 
so $x$ is $O(x)$ and $O(3x)$ and $O(0.01x)$
 
right
like XeR and YeR
 
Ugh these headshots
 
not sure what the existential quantifier with the naught n is either
 
2:52 AM
So, let $f(n) = \frac{1}{n} +15 - 3$.
 
but I should be able to solve it this way
 
@ErikHumphrey for instance $x+5$ is $O(x^2)$ because $x+5\le x^2$ for all $x$ past a certain point (that's the naught thing)
 
what are your favourite theories?
 
vitalism and aether
 
mathematical theories
 
2:54 AM
$f(n) = \frac{1}{n} + 12$. If we want to show $f(n) = O(n^3)$, then we have to find an $n_0$ and $M$.
 
What counts as "a theory"?
 
can we not
 
@Daminark a language is a set of symbols with constants, functions, relations, logical symbols, variables, and puncuations
a theory is a set of sentences over said language
 
my dad sees me reading a book on Galois Theory "Remember, it's just a theory."
10
 
$\frac{1}{n} + 2 \leq n^2 + 12n^3$
 
2:55 AM
Repeat multiple times
 
where a sentence is a well-formed formula with no free variable
 
@AkivaWeinberger LOL
 
could I simplify this further?
I think that's all I need to solve it
 
and I won't even try to define "well-formed formula"
 
Okay better question, what's your favorite theory?
 
2:55 AM
Now I can tell whether or not it's O(n^3) by entering small numbers and absurdly large ones
 
Yeah, sure, you're done i guess... I would've preferred you showing its O(1) then you get everything else for free
Uhh
 
@Daminark I think a theory is good iff it is complete, so examples are torsion-free divisible abelian groups, presburger arithmetic, dense linear-ordering without endpoints, and algebraically-closed field of fixed characteristic
 
I don't know how to show a function is O(1)
 
@ErikHumphrey where are you getting $n^2+12n^3$ from? do you mean $1/n+12$? on the left?
 
One point of a proof is so you don't have to do any more guessing and checking.
 
2:56 AM
@Daminark jokes aside, I think Galois theory is quite cool
so is category theory
 
$\frac{n^3}{n} = n^2$ yes?
 
in any case if $n>1$ then $A\le B$ implies $An^3\le Bn^3$ so indeed $1/n+12\le (1/n+12)n^3$ if that's what you mean
 
sorry fixed
 
$\frac{1}{n} + 12 < 2 + 12 = 14$. Done.
 
why 2?
 
2:57 AM
$12n^3 = 15n^3 - 3n^3$
 
@anon because $1/1 = 1$
 
Why not?
 
Oh, I'd usually refer to that as a field of math more than a theory. The main things I've heard referred to as theories are the cohomology theories
 
n2 because n3 / n = n2
 
@LeakyNun oh, you made it a strict inequality
 
2:58 AM
@anon I didn't
 
or rather, orbit did
 
in this case I believe it is not a proof but rather some kind of non-proof derivation
 
Yeah. Oops. I go back and forth between the two
 
I'm not the best with actual proofs either
let me input some values
 
@ErikHumphrey ??
 
2:59 AM
Number theory is probably my favorite thing I've interacted with
 
ew
@Daminark
 
Abstract Algebra is the best thing ever
 
oh THAT 2
 
Lol @orbit, what's your favorite thing?
 
2:59 AM
@MatheinBoulomenos abelian groups are nice
 
I haven't found it yet :( But I know it's not number theory
 

« first day (2675 days earlier)      last day (2345 days later) »