« first day (2450 days earlier)      last day (2867 days later) » 

00:00
There is a professor here who (I heard) he was instrumental in getting to go to grad school, after a few years doing things not directly related to math after being a prodigy type undergrad student.
That was a tough sentence to parse.
I tried.
deT@, eno siht gnisrap yrT
eyes bleed
Sometimes Zach acts like he's ... um ... 14 years old.
00:03
Well anyway, see you around everyone! Talk is starting now
Lol @Ted
Bye, Demonark.
I'm not that crazy
takes bets
Whats the answer to my probability question though?
Did anyone look at that?
00:07
Say you start with $n$ dollars, and you start playing a game where you can either win 1 dollar or lose 1 dollar. Say you win with probability $p$. What is the probability you lose all your money approach after playing this game a number of times going to infinity?
I'll solve that for $n=1$.
Actually, there are good ways of doing this which I surely knew when I taught probability.
Yeah, you have to use the "fractal" or "self-similarity" approach.
But I'm not going to think about it now. Have to go play bridge tonight.
Good luck with your bridges
LOL
I'm gonna play poker... Hopefully I won't be poked too much.
Nate has returned...
See you back at the math games sometime, Zach :)
The math games?!
You know ... where we try to do math ?
00:15
@MeowMix Actually the reason I was thinking about my question was for bankroll management...
00:49
I am @TedShifrin !!!!!
01:10
can some one point me in the right direction how i can show 3/(7x-4) = 1/( 1+7/3(x-1) )
01:53
So here's a question. What are some theorems whose statements are pretty short and elementary, but whose shortest known proofs are really large and complicated?
The four-color theorem is sort of the prototypical example. Pretty simple statement, extremely complicated proof.
Maybe there's some simple Diophantine equation where you need complex analysis to prove that it has no solutions?
Oh right, Fermat's last theorem is that kind of thing. Course, saying that that proof "needs complex analysis" is a major understatement.
If the Goldbach and Collatz conjectures are ever proven, they'll probably be good examples.
02:11
@TannerSwett The proof for fermats last theorem really fits well. It is such a deep and difficult result (modularity theorem)
Hey everyone!
Salam @Ali! Long time no see
Hi @Daminark I have been studying for my finals, but I am being naughty tonight
What have you been up to
Just going forward with classes and all
Anything interesting?
Well, this quarter in analysis we're doing measure theory + miscellaneous topics (e.g. Banach Mazur game), and I'm also taking differential topology
02:25
4
Q: A minimization problem in function fitting setup

Rajesh DachirajuLet $\Omega$ be a convex, closed, compact set in $\mathbb{R}^d$ with a smooth boundary. Given a data $(x_i,d_i)$, $x_i \in \Omega$,$d_i \in \mathbb{R}$, $i = 1,2,3...N$, and $\sum\limits_{i=1}^N d_i = 0$ I want to find a function $f:\Omega \to \mathbb{R}$, such that, $\int_{\Omega}f(x)dx = ...

oo Differential topology certainly has lots of interesting motivations
There is a large motiation from physics as well
Yeah, it's been pretty fun so far. This last class we showed the continuous and smooth homotopies were, in fact, the same thing, and now we're starting on the degree of a map
Degree stuff is great
You can prove things like hairy ball theorem
We're gonna do that pretty soon, I think
Thats very fun
It makes you think how weird it is that you can have a tangent field on s^3 with no cowlicks
but alas I am constrained to analysis 1 and linear algebra for now
How has your measure theory course been like @Daminark?
02:30
It's been pretty nice
The first part of the class is probably best described as descriptive set theory
Banach-Tarski, then sigma algebras, Borel sets (proving that the hierarchy doesn't terminate in countably many steps), Banach Mazur
We only reached measure theory a few days ago :P
Caratheodory extension, Lebesgue measure (few proofs because details were boring), and now the Lebesgue integral
My only measure theory was Lebesgue integral and some probability
but apart from that I know literally nothing
Its really difficult to find informal overviews of the subject
TerryTaos blog posts seem to be the best I found
Banach-Tarski is again another interesting result
but maybe I am just not cut out to be an analyst
I tend to find the proofs quite monotonous
or I am extremely bias towards algebra
Hey anon
Well, the problems my professor is giving us are less, go into long tedious detail, and more difficult conceptual questions that have a trick
But I do realize that analysis on the whole definitely has a feel of, take something that should morally speaking be true, and then have an arduous $\epsilon$ proof in order to hash it out properly
Which is why I anticipate that I'm gonna like algebra at least a little bit more, it feels more like the work is in understanding stuff and sometimes has more of a puzzle feel to it
Though as of now I'm in the "kind of but not really ish know what's sorta going on in group theory maybe??" plane of existence
Next year I'll take a whole lot of math classes and actually figure out my life
Yeah, I think the thing to remember is that you can always change your field if you don't like it
Its not going to be set in stone if you choose this or that
02:50
Yeah, definitely
I'll just make sure that wherever I go for grad school, it has at least some decent representation in various fields
I'm really hoping for Michigan
But yeah Peter May was actually giving a talk on applying for grad school and all
03:18
@TannerSwett You might find this old question useful: mathoverflow.net/q/51531/55904
 
1 hour later…
04:21
@arctictern Can I ask you a question? I am sorry its a stupid question.
@arctictern I have a map $\alpha: [0,1] \rightarrow X$ and $e: [0,1] \rightarrow Y$
Does it make sense to define $ \alpha \circ e^{-1}: Y \rightarrow X$?
what do you think?
I dont think it does unless the $e$ is invertibe
invertible
pretty much
04:27
Is there any way to force this to happen?
one could make sense of $\alpha\circ e^{-1}$ without $e$ being invertible, as long as (i) $e$ is onto and (ii) $e(t)=e(s)$ implies $\alpha(t)=\alpha(s)$.
@Jaynot huh?
Yes you are right @arctictern. In a case in $ e: [0,1] \rightarrow \triangle^1$ such that $$ e(s)= (1-s)\underline{0} +s \underline{1}$$ and let $X$ be an arbitrary space. Can I do that?
*In a case in which
dunno what $\triangle^1$ is
Oh sorry, its just 1-simplex
so $e$ is invertible in that case
04:36
@arctictern. Wow thats really great. I actually wrote a proof to a problem but I was worried its not invertible. Yeah I am good
@arctictern but wait why is it invertible?
a 1-simplex is just [0,1] innit?
Yeah
what are $\underline{0}$ and $\underline{1}$, that you can multiply them by real numbers and then sum?
@Jaynot $e$ must be surjective for $e^{-1}$ to be a function with domain $Y$; $e$ must be injective for $e^{-1}$ to be a function; so $e$ must be bijective.
@arctictern If $e$ is not invertible, then how do you form $e^{-1}$ in order to form $\alpha \circ e^{-1}$
one doesn't; one writes $\alpha\circ e^{-1}$ as abuse of notation in that case. (you can interpret $e^{-1}$ as taking the preimage if you want)
@arctictern ...
Preimage of e, is e the identity?
@Secret no.
25 mins ago, by Jaynot
@arctictern I have a map $\alpha: [0,1] \rightarrow X$ and $e: [0,1] \rightarrow Y$
e, that's 2.718 right guise?
@arctictern $\underline{0}$ and $\underline{1}$ are affinely independent
04:47
24 mins ago, by Jaynot
Does it make sense to define $ \alpha \circ e^{-1}: Y \rightarrow X$?
@Jaynot so, verify that $e$ is injective and surjective
Ok. in between this actually the question I am working on math.stackexchange.com/questions/2239895/…
@arctictern
Okay, I found $e$ to be injective. Setting $e(s_1) =e(s_2)$, we have that $ -s_1 \underline{0}+s_1 \underline{1} = -s_2 \underline{0}+s_2 \underline{1}$ which implies that $s_1 =s_2$ by comparing coefficients.
and e is also surjective too.
@DHMO I am considering defining the map that way based on the way the question was posed math.stackexchange.com/questions/2239895/…
05:16
Hey guys! I've been working on this for a while now. I can't seem to express this well in terms of a formal proof
$f(y) \geq f(x) + <\nabla f(x),y-x>$ if f convex and once differentiable
I see that it's pretty obvious - you draw a tangent line and it's lower than any point except at x
Taylor series would be a nice way to motivate this but ultimately I want to explain it by properties of convexity
$\langle a \rangle$ \langle a \rangle
thanks for the reference - I'm just a little lazy
I'm sure it gets the point across well enough
Hi Ray
Hey Mike! What's up
nevermind - solved it
dislike tricky algebraic nonsense manipulation though
05:36
that's what you were asking for, wasn't it? :)
Well if by some dumb luck I missed something very obvious then maybe it wouldn't be so ugly :<
You should be able to prove that convexity is equivalent to "the graph lies above the tangent plane at every point", and that your inequality is exactly the same as "the graph lies above the tangent plane at $x$"
The first part is gonna be algebra though
I see. Yeah, I pretty much just took the definition of $f(\alpha x + (1-\alpha)y)$ and took $\alpha$ to 0 and got the definition of the gradient, which is pretty much what you said :)
Yup, good job.
164 is an interesting class thumbs up disclaimer: in the eyes of a cs major
05:41
What class is that? Optimization?
yes, taking it with Wotao Yin
gotcha
not my game
Yeah I remembered to put the disclaimer down haha
Hey everyone!
@OneRaynyDay Is it that you're a CS major and happen to like that class? Or is that a class in compsci?
05:48
I thought of it as more of a stats class, at least it's in that department where I'm at
Hey @Alessandro!
it's a math class
he's a CS major
@Daminark I like it because it's very important to applied situations, where CS is applicable thumbs up
hi everyone
Ah, I see
Hey @Balarka!
05:50
And lol to be fair optimization feels more like math than stats for sure
for example in a lot of ML papers, they use strong convexity to prove stuff like new optimization methods converge faster than normal stuff
I know nothing of stats except for the minimal basics so take anything I say with a grain of salt, but yeah
and i mean just optimization in general is just used everywhere. And yeah stats is not really involved here
Ah, that's neat
except if you wanna add stochasticity and stuff into your iterative updates or something :o
05:53
Lmao, perhaps
shrugs
@Mike You mentioned some time back that the type of number theory you were into wasn't as well represented in UCLA as you would've liked
How would you, in general, divvy up the topics by how they're represented there?
"1. heegaard floer homology" ? :P
@Balarka Is that actually a thing? I'm guessing no
ya it is
But I was just joking.
And lol I meant more general than that, just like, for example algebra/algebraic geometry is well-represented, p-adic analysis and IUT are most common, analysis/PDE less so, etc
IUT = Inter-Universal Teichmuller theory? That's common???
06:03
I was joking as well
it's a really big dept
i'd rather not enumerate what people do, at least not without just being able to vocalize it in person
the number theory thing was really: I wanted to count rational points on curves (equivalently, I wanted to find rational/integer solutions to polynomial equations - are there any integers with $x^2 + y^3 = z^7$? which ones?), and here people are more modular formsy
Ah
That's good that at least there's a lot going on
I don't think as an undergrad I'll be comfortable having more than a vague sense of wanting to go one direction or another, so I'd definitely hope to look at broader places, instead of like, powerhouse in topos theory and then little else
I've seen a lot of stuff that I've found to be cool, to the point where the desires feel almost local. Right now I'm totally gonna be into complexity theory, tomorrow algebraic geometry, next day next random subject
@Daminark In my second year of undergrad I was learning what a group was. You'll be fine. You'll either have a good feel of what you want to do by then or you won't, and you'll apply to grad school appropriately.
Do most math majors apply to grad school?
(fwiw, if I went to a school that did number theory closer to what I did, like Brown or MIT, I think I would probably have ended up doing that)
No
06:15
I see - what's the typical track then?
Oh really?
The vast majority of people don't plan to stay in academia (and shouldn't!! there aren't enough jobs and it can be pretty miserable if your field isn't, like, your life dream) - I think a lot of people go straight to studying eg stats or finance
that's what happened at my ugrad
I see. I may be a bit biased here but as I'm living in the communal dorms in UCLA(on the hill)
Here I've heard that somewhere over a half of students try for grad school? Though not necessarily math
06:18
I see people who majored in statistics or finance kind of want to "take it easy" and are usually in fraternities or do not really like the field as much, mixed with the extremely passionate stats people
Stats and compsci seem to be the next contenders, followed by physics
However, for those who are in mathematics and plan on staying that way usually take a long time thinking about equations and proofs, and are more absorbed by math in life
But I'd say that the stats undergrad department in UCLA is a bit "diluted" in comparison
I don't actually know much about the stats department here tbh
I can look at the ugrad population I've seen at UCLA vis a vis teaching 132 or whatever and I can honestly say that the vast majority of people taking that class should not try for grad school
The number of people who actually understand what's going on is a minority
My guess is around 5-10 or so people who should "really go to math grad school" graduate from UCLA each year
The math major here requires, for some reason, that people basically take 4 (or 7 for the BS) classes outside the math department but within the physical science division, and math majors who don't have particular interests in some other subject tend to go for stats or compsci as being closest
How many math majors are there total?
06:22
a lot
I'm asking other people for guesstimates
I agree with Mike's sentiment, but to the guess of 5-10 I'm not sure
That sounds like an awfully small amount of people :o
meh i cut my toenail too deep and it's a bloody mess now
@OneRaynyDay I'm holding a pretty high bar here. Maybe a little more. I'm basically trying to count the number of people who take grad courses before they graduate.
If you're doing your ugrad at a school that offers those, and you want to go to math grad school, you should probably start taking them during your ugrad
06:28
I am in pretty bad position to judge people's preparedness of course, and it's hard to use the numbers of how many people go for it since some people seem to end up stumbling on grad school, while others end up finding it's not for them and go to either tech or consulting
I see. That's a good metric I suppose
But out of 150 math majors yearly, about half at least try to go to grad school
I'm guessing you're counting those people who do putnam in ucla
Ah, by your metric the number would likely decrease
My guess is about 100 pure math majors and about 10 of them go to an R1 school (approximately top 50) for a pure math PhD program (more go to applied)
My number drops by only counting straight to PhD (not people who go to a terminal masters)
06:32
Lol so Greg Lawler was at the talk along with Peter May about getting ready for grad school
Many of us were amused that he was definitely one to be like "Don't forget stuff like probability and all"
I mean, the way he put it was probably fair. Biased, but kinda correct
It might be that Chicago has a much stronger ugrad-to-grad-school pipeline
He often also took the role of, guys remember that academic jobs are tight and the whole system is kind of unstable, so you should have some backup plan lest things go sour. And made sure to emphasize that you get a fellowship of some kind
Maybe I'm the Lawler here
You need to remember that and talk to other students about it
If you're paying for any aspect of grad school out of pocket DO NOT GO
The only situation I can think of where that's not accurate is if you're really really committed to the PhD path and you want to get a terminal masters (eg Cambridge Part III) to boost your resume
Perhaps, I'd attribute it partially because grad school just sorta feels like the natural next step, some might not necessarily intend to try for academia afterwards but hey, math was fun now, it'll be fun then, and I can figure something else out afterwards
I'm still trying to see whether or not I fall into that category
Yeah I hear you
Ring $R$ polynomial $p$ reducible to linear factors => it has roots?
I know this is false, but why?
and why is it true in field $F$?
06:45
Why is this false? If you can write $f(x)=(x-a)g(x)$ it has $a$ has root?
@AlessandroCodenotti consider $4x^2-1$ in $\Bbb Z[x]$
I think maybe DHMO means in the opposite direction?
@BalarkaSen if it has roots then of course it can be reduced to linear factors
No
Only if integral domain
Eh, nah.
so neither the converse nor the positive is true? @BalarkaSen
could you provide me with an example?
06:48
@DHMO that has no roots in $\Bbb Z$
@AlessandroCodenotti exactly
@Alessandro But it does factorize.
I seriously have no idea how the Veblen function works (and I suspect that $\psi_{\beta}(\alpha)$ in $C_{\beta}(\alpha)_n$ is not the veblen
**Tier 0**

$0=\emptyset$

**Tier 1**

$1=\omega^0=\{0\}$

$k=\{0,...k-1\}$

$\omega=\sup(\{0,1,2,...\})$

**Tier 2**

$\omega+1=\{j,\omega\vert j\in\Bbb{N}\}$

$\omega+k=\{j,\omega,...,\omega+(k-1)\vert j\in\Bbb{N}\}$

$\omega + \omega=\omega 2=\sup(\{j,\omega,\omega+1,\omega+2,...,\vert j\in\Bbb{N}\})$

**Tier 3**

$\omega 2+1=\{j,\omega+j,\omega 2\vert j\in\Bbb{N}\}$

$\omega k=\{j,\omega+j,\omega 2,...,\omega(k-1),...\vert j\in\Bbb{N}\}$

$\omega \omega=\omega ^2=\sup(\{j,\omega+j,\omega 2,\omega 3,\omega 4,...,\vert j\in\Bbb{N}\})$
@DHMO No, I was misremembering. Long division should work everywhere.
@BalarkaSen ok
06:49
Wait really?
@BalarkaSen well the converse is really the factor theorem
which I never bothered to prove
Yeah, you can definitely long divide. It might not be unique however.
@BalarkaSen ah, right, because not everything can be assumed with leadinf coefficent 1outside a field
meh, algebra is too confusing. i back off from this
Wait hold on I'm still not buying this
So the problem with rings is that unless they're integral domains, non-zero things can multiply to $0$
06:53
The factor theorem is based on the division theorem
If you can factor out $f(x) = (x-a)g(x)$, I was under the impression that $f(a) = 0$ immediately. Why is this not the case?
@Daminark this is true
but you should read my "theorem" again
15 mins ago, by DHMO
Ring $R$ polynomial $p$ reducible to linear factors => it has roots?
linear factor $\ne$ monic linear factor
You can always divide by a polynomial whose leading coefficent is a unit even if you're not in an euclidean domain I guess
Oh and you can't divide out in rings, right that messes it up
In particular if $a$ is a root division by $(x-a)$ should work
06:55
@AlessandroCodenotti what are you trying to argue?
Yeah the thing that stuck out to me more was that since you were not an integral domain, hypothetically $f(a) = 0$ while you couldn't guarantee that it was the product of things that were $0$
That division should work even if you're not in an integral domain to get linear factors
@Daminark good point
So that's why I was thinking that what you said was true while the converse was false
But I'm not convinced, take a secons degree polynomial with 4 roots, dividing by all 4 of them should give strange results
06:57
Lol I should like, actually learn algebra
can someone give me a non-integral domain to work with?
Before debating this stuff
Z/nZ with n not prime
@DHMO Z/6
thanks
that's nicer than I thought
so $2x-2=0$ in $\Bbb Z_6$ has two roots
namely, $1$ and $4$
06:59
Lol I've heard that in ring theory, there are a lot of things people want to be true that aren't, because everyone seems to always think of just $\mathbb{Z}$ and $\mathbb{F}[x]$ as the examples of anything ever
@Daminark I think $\Bbb Z_4$ basically rules out any nice things
Tier 0 is doing nothing $\emptyset \mapsto \emptyset$
Tier 1 is addition $k \mapsto k+1$
Tier 2 is transfinite addition $\omega+k \mapsto \omega+(k+1)$
Tier 3 is transfinite multiplication $\omega k \mapsto \omega (k+1)$
Tier 4 is transfinite exponentiation $\omega^k \mapsto \omega^{k+1}$
Tier 5 is transfinite tetration ${}^k\omega \mapsto {}^{k+1}\omega$
Tier 6 is epsilon exponentiation $\epsilon_k \mapsto \epsilon_{k+1}$
Tier 7 is a tower of epsilons $\psi_?(?) \mapsto \psi_?(?)$

and then I am stuck at Tier 7 because I have no idea how to write $\epsilon_{\epsilon_{\ddots_{\epsilon}}}$ w
what are some properties with $\Bbb Z_n$ with composite $n$?
@AlessandroCodenotti here's the weird thing
$2(x-1) \equiv 2(x-4)$
but $x-1 \not\equiv x-4$
so $\Bbb Z_6[X]$ is not a UFD, @BalarkaSen right?
That's not weird, you can't conclude $b=c$ fron $ab=ac$ in a non integral domain
UFD are assumed to be integral domains
unique factorization domains are automatically integral domains
07:05
I mean I'd probably peg it as one of those things whose truth is easily acknowledgeable, but which sounds like witchcraft the first time around
can anyone kindly provide me with a non-ID that whose only unit is trivial?
I think I need to discuss more with @SimplyBeautifulArt about the various Tiers of the ordinals cause starting from the Veblen, it gets confusing

The following is what I knew regarding fixed points:

Tier 0: $x=x$
Tier 1: $x+1=x$
Tier 2: $x+\omega=x$
Tier 3: $\omega x=x$
Tier 4: $\omega^x=x$
Tier 5: ${}^x\omega=x$
Tier 6: $\epsilon_x=x$
Tier 7 should be $\psi_0(x)=x$ ?
and Tier 8 should be $\psi_x(0)=x$ ?
Therefore the first fixed point of every tier are:
Tier 0: 0
Tier 1: $\omega$
Tier 2: $\omega 2$
Tier 3: $\omega^2$
Tier 4: ${}^2\omega$
Tier 5: $\epsilon_0$
Tier 6: $\epsilon_{\omega}$
Tier 7: $\zeta_0=\psi_1(0)$
Tier 8: $\Gamma_0$
Tier 9: $\psi'$
Tier 10: ???
Good lord
And @DHMO I'll think about it, but I prob won't be of much help, I know few examples of this stuff
alright thanks
07:24
Actually wait @DHMO you can have non-trivial units in integral domains as well
yes we can
$\Bbb Z$ has $-1$ as a unit as well
Dear lord this will be tricky
$\Bbb Z$ is an integral domain
Yeah that's what I'm saying
Any $\Bbb Z_n$ has many units
as many as $\phi(n)$
(can you prove it?)
07:26
I'm assuming $\phi(n)$ is the number of factors?
no
In number theory, Euler's totient function counts the positive integers up to a given integer n that are relatively prime to n. It is written using the Greek letter phi as φ(n) or ϕ(n), and may also be called Euler's phi function. It can be defined more formally as the number of integers k in the range 1 ≤ k ≤ n for which the greatest common divisor gcd(n, k) is equal to 1. The integers k of this form are sometimes referred to as totatives of n. For example, the totatives of n = 9 are the six numbers 1, 2, 4, 5, 7 and 8. They are all relatively prime to 9, but the other three numbers in this range...
It is the number of positive integers less than $n$ and coprime with $n$
@DHMO $(\Bbb Z/2)^2$
or ^n
@MikeMiller thanks
what are the group operations?
That makes sense, since if $m$ is coprime with $n$, you can raise it to powers and never get $0$, so you raise it to a high enough power that you get 1
(Do let me know if I'm spouting nonsense, at 2:30 my ability to think degenerates)
07:30
componentwise addition
everything is componentwise
(Not precisely 2:30 but like, at night. I should get some sleep eventually)
@Daminark do you know what a unit is?
Basically something which you can invert
@MikeMiller so every element is order 2
07:32
@Daminark yes
so how does it matter if you can get $1$ by raising it to a high enough power?
Well let's say you have $m^k = 1$
Then $m^{k-1} = m^{-1}$
@Daminark is $3$ coprime with $6$?
I'm asking you a question.
I didn't realize you were asking a question
They're not
It ends with a question mark.
I've seen a lot of counter statements written in question form
07:35
@Daminark and which power of $3$ in $\Bbb Z_6$ is $0$?
... Crap I'm dysfunctional
@Daminark sleep
@MikeMiller Is $(\Bbb Z/2\Bbb Z)^n$ isomorphic to the powerset of an $n$ element set with symmetric difference as addition and intersection as multiplication?
(That's the example I was thinking about for DHMO's question)
sounds about right
to me too
because pointwise addition and multiplication in $\Bbb Z_2^n$ is the same as XOR and AND
which also applies to powerset of an $n$ element set
since the powerset can be represented as $n$-bit binary numbers
07:40
Oh wait right you can do the thing with a combination equaling 1
Like if $(x,n) = 1$, then $ax + bn = 1$ for some $a$ and $b$, so that $ax = 1$ in $\mathbb{Z}_n$
And this should work both ways
@Daminark what theorem is that?
I think it's called Bezout's identity?
yes
@BalarkaSen what is the theorem that $U(6) \cong U(2) \bigoplus U(3)$?
Nice, so that works. Alright, marginal feeling of competency restored
@DHMO Chinese remainder
07:44
@BalarkaSen ...
But it's easier than that.
the ellipses constitute neither a statement nor a question so i'm going to pretend it does not exist
3
Well, I should probably go to bed at this point, lest I start saying that $S_n$ is simple or something
See you guys around!
Good night.

« first day (2450 days earlier)      last day (2867 days later) »