« first day (4117 days earlier)      last day (1199 days later) » 

00:14
So it would seem that integrating floored logarithm gives a nice approximation of the integral of the logarithm. It also seems to be absolutely precise to arbitrary precision for a domain $\frac{\lfloor 2x\rfloor}{2} \in\mathbb{N}$ for some reason. Probably something to do with the built-in functions being used.
Not sure how to properly describe that behavior though.
The error is also incredibly regular with the error being a waveform whose local maxima increase linearly as x approaches positive infinity.
That's not why I'm here though. Does anyone have any ideas on how I could get back to an approximation of $\log_2(x)$ without having to divide by $x$?
The integral that I'm approximating is $\frac{x\ln\left(x\right)-\ln\left(2\sqrt{2}\right)x}{\ln\left(2\right)}$.
Apologies for the mess, but here's the workspace for context: desmos.com/calculator/senxrt2b0i
By the way, I found this excellent maths channel on YT: youtube.com/channel/UCipYk6snJe0IaOvzSCWwTrw
Not sure if it helps, but I learned a while ago that there's a handy search function built into desmos, so you can just search latex strings directly (as formatted by desmos).
00:32
how does one pronounce desmos. demos? dess-moss? what. they should have thought about this.
I pronounce it "dehz-mohs"
Very greek-sounding
(Not to be confused with [Kalle] Demos :P )
I just need this and my exp2 approximation to be accurate to 5 decimals and I can be happy since then I only need two iterations of Newton-Raphson (which I learned that the recursive formula I found was just a rediscovery of Newton-Raphson reciprocal approximation).
I also learned that wolfram can properly solve for explicit formulas.
they do appear to pronounce it dez-mohs
some people soften the z to an s
i really think this should be on their main page
there's a reason why a lot of international brands are easy to pronounce, they are getting in their own way here
The explicit formula for NR reciprocal is $\frac{1-(1-xy)^{2^n}}{x}$ for iteration $n$ and initial approximation to $\frac{1}{x}$ as $y$. Pretty neat!
So in theory, if you could cheaply compute $\frac{1-xy}{x}$, one iteration of NR is sufficient to compute reciprocals to arbitrary precision.
err zero iterations
In any case, such magic does not exist in this world which is why I'm looking for a way to compute this fast log2 function (getting the fast exp2 was easy enough by once again integrating the exp2 approximation itself).
Any ideas?
Maybe there's a way to work directly with the (indefinite) integral approximation of log2?
no ideas here.
here meaning specifically me. not here here.
Unfortunately I can't use the limit definition of the derivative of a function to approximate the log2 approximation here as it will just approximate the log2 approximation that I integrated.
And there's this one pesky term that doesn't have an x that I can eliminate, hence the issue.
Namely the term $-2^{\operatorname{floor}\left(\log_{2}\left(x\right)\right)+1}$
That is, unless I can manage to weasel out the $x$ there as a coefficient which seems plausible but not too obvious.
I mean by all intuition $-2^{\operatorname{floor}\left(\log_{2}\left(x\right)\right)+1}$ is just a variation of $\lfloor x\rfloor$ in one form or another.
I appreciate the conversation anyways, Leslie :)
What do you think here of this thing having the capacity to compute exact multiples of radical two? You can see for yourself in the graph.
Also, can I call this waveform sinusoidal or is that reserved for waveforms strictly derived from sine or cosine?
00:50
@leslietownes Years ago, I ran into a bunch of Desmos developers in an airport (they were talking about mathematics, I was eavesdropping, then inserted myself into their conversation). They pronounced it DEZ-mohs.
I don't think it can get any more authoritative on pronunciation than that!
there should be a FAQ page linked on their main page which says exactly this.
or better yet a name that didn't require explanation. i should apply for a job as a branding consultant.
Yeah, wouldn't want some millennial memelord calling Desmos Deez mohs.
@AMDG I pronounce it "deez nutz".
ecks dee
00:52
see? i could have told them to avoid this.
err, I mean, ok boomer. Gottem.
Yesterday I learned that a lagrange interpolation of log2 effortlessly beats a minimax maclaurin approximation of order 2 effortlessly.
(With zero coefficients, same order 2, uses three polynomials)
Well, apart from the coefficients that normalize the polynomials to 1
Approximation theory is spicy and I love it
01:25
Hm, is there a way to represent the natural logarithm in terms of an inverse of the exponential as $\cosh(x) + \sinh(x)$?
according to some, the natural logarithm is the inverse of the exponential. i haven't looked into this.
lol no I mean is there an identity of the natural logarithm that explicitly makes use of hyperbolic functions or otherwise?
probably, in that people often define those things in terms of the exponential. but you seem to have some numerical goal in mind, i dunno about any of that.
the festival of birthdays is coming to a close, unless anyone has a birthday tomorrow.
Oh wait, I just remembered an identity of arsinh
So $\operatorname{arsinh}\left(\frac{x^{2}-1}{2x}\right)=\ln\left(x\right)$. Found what I'm looking for.
01:35
i thought the diner that i ordered my friend's birthday gift from had an exorbitant delivery fee but i was wrong about the units. i thought i was ordering a slice of pie, and they gave her an entire pie.
symbolic integration packages often kick out this kind of stuff.
You functional analysts never were any good at units.
this is true. at least i was off only by a fraction of $\pi$ and not off by $2\pi$ (ha, ha).
smack
Tell munchkin her turn to smack you!
somebody hit her yesterday at school and didn't apologize after. we haven't gotten to the bottom of this.
Oh oh
Was she being particularly bratty?
01:39
we know his name is "al" and it occurred on the playground. she was using something al wanted. that's all we've gotten.
it sounded like she had a legitimate grievance. i'm very thankful that whatever faults she may have, hitting kids is not one of them.
Sounds like thuggery. A rethugnican politician in the making.
i was talking to a friend today. her son, just younger than my daughter, has a habit of hitting people and throwing sand on them on the playground.
i wouldn't know where to begin with that.
Too. Much. Violence.
i'm deliberately not following the trial of that guy. you know the guy.
Yes, America is grand.
Did you get me my gin ?
01:41
my daughter's main faults are yelling about her imaginary friends at the wrong times, and (sometimes) saying that people are poo-poo if they don't do what she wants.
i forgot. let's say it's in the mail.
Uh huh. Screech is still growing ….
if it wouldn't lead to violence, which it would, i do think a play date for screech and olivia would be a good idea. our cat needs to learn that there are other cats and she isn't the center of the universe.
A lesson for both.
she's on her thermo kitty bed right now, running up the electricity bill.
Screech is lying on the arm of the couch looking like a bat.
Only thermo she gets is the covers on my bed.
01:47
i assume she exercises her claws on the couch. if not, that's something she might consider doing.
we have two sheepskin rugs we threw over the arms of the couch to keep her from scratching on them. she now jumps on top of the couch and scratches the cushions instead of the arm rests.
sheepskin at right.
Narcissistic kitty.
we have to lint-roll her fur off of that thing every day. it's good that we got a white couch.
one weird thing i've noticed somewhat lately is people downvoting questions that comply with site guidelines but are, for lack of a better word, kind of dumb. this seems to be missing the point. if someone can clearly lay out how they are thinking wrongly about something, maybe worth an upvote or just leaving it alone.
anyway, it's clearly not 'do my homework' if someone is very up front about how wrong they are about something. the verbalization of the wrongness reflects thought.
I typically only downvote wrong answers when the answerer won’t budge after comments.
01:56
i think sometimes people use downvotes for 'i don't want to see more of this,' which i get, but, doesn't seem to be what the site is set up to do.
I vote to close way more than I downvote.
same. i've only downvoted two or three times. in each case it was someone going out of their way to be insulting.
i'm generally impressed with the level of discourse on main. i tend not to engage when i feel like someone might not respond appropriately and if i did i would probably feel differently about the whole site.
most of the 'do my homework' folks are appropriately ignored/closed, it kinda works. it's not a disaster. that's all that the internet can aspire to.
02:12
Ok, I managed to find a solution to my logarithm problem. It's two decimals of precision like the lagrange approximation, however, no multiplies are strictly needed afaik
Improving the accuracy of my exp2 approximation consequently also improves the log2 approximation since the integral is computed using exp2
Yep, the only multiplies in subsequent integrations are multiplications by constant coefficients, not $x$.
Noice
So the cost is at worst O(M(n))
Will clean up and then post here for anyone interested.
Boy this day couldn't get any better though...
Using the second antiderivative of exp2 approx. and first antiderivative of log2 approx., the error for computing reciprocal is less than 1e-5 after x=17 or if only using the first antiderivative of the exp2 approx. and first antiderivative of the log2 approx., it is less than 1e-5 after x=61
Using a table of 16-bit reciprocals with only 18 or 62 integers is only 36 or 124 bytes respectively.
Actually according to the absolute error I need slightly more integers than those mentioned but it's still way better than only less than 1e-5 at x=2000.
02:42
There we go. Clean workspace: desmos.com/calculator/zs7rjdcpj7
Hi chat!
I have been playing around applying integrating factors to different PDE forms, and it seems to me that they can be used to solve $au_x + bu_y + cu = 0$, where $a$, $b$, and $c$ can all depend on $x$, or all depend on $y$, but not both. So they can solve "collectively single-variable" coefficient homogeneous linear PDEs. Is this as general of PDE as the technique works for? Or is there some combination of $a$, $b$, and $c$ depending on both variables?
I am having a transcendental equation, say $f(x) = x^2 e^{Ax + B} + C + Dx + Ex^3$ and I am trying to determine the number of roots it has witht he variation of $A,B,C,D,E$..I was playing with desmos and could see that it can have either one root or three roots
real roots
Also, I take it this doesn't work at all for PDEs in $3$ independent variables, unless you get lucky?
I am trying to find when the number of roots changes like a condition?
any ideas??
02:54
Maybe start by noting two similar variations that yield a different number and create slider(s) in Desmos for the variables that differ between the two. Then you will be able to see how one geometry changes into the other, and pinpoint the in-between state.
You could also set sliders for all of them at different speeds, and just watch until something interesting happens.
03:09
@user10478 this sounds like method of characteristics stuff
In the same ballpark yeah. Maybe the integrating factor technique is just superfluous since the method of characteristics solves more problems.
I wouldn’t call it superfluous, if you can use it to solve a given problem more simply
But characteristics may indeed cover more cases
03:51
@BalarkaSen I heard that, in Hindi, the names for the numbers from 1 to 100 are so irregular, you basically have to memorize them. Is that true?
 
2 hours later…
05:56
Hello; I have a question about math.stackexchange.com/questions/3871966/… specifically the last sentence of the answer which says " it means that we get terminating expansions, rather than infinite strings of 1s in our binary expansions " isn't this untrue?
 
1 hour later…
07:09
for example $\frac12=0.1\bar0_2=0.0\bar1_2$
07:55
^IGNORE I didn't fully read the answer its talking about the expansion constructed in the answer
08:34
HI Everyone! What is the intersection of this two sets?
$\biggl\{ \norm{X_n-\mu} < \frac{\delta}{2} \biggr\} \cap \biggl\{ \norm{X-\mu} < \frac{\delta}{2} \biggr\}$
sorry I messed up the formating
$\biggl\{ ||{X_n-\mu} < \frac{\delta}{2}|| \biggr\} \cap \biggl\{ ||{X-\mu} < \frac{\delta}{2} || \biggr\}$
that's what I meant, sorry
I'm trying to show that $ \{||{X_n - X} \geq \delta || \} $ is not in it
09:09
My whole attempt: imgur.com/a/0J60cpd Ant tbh I don't how to proceed
10:08
nvm, I solved it
 
1 hour later…
11:33
Are there ways to solve for n in something like 3^n - 23^n= -6?. In general, stuff like a^n - b^n = c? Without numerical methods I mean
11:46
Hello I have questions in mathematical physics
why there is cos alpha in Integral?
12:09
I understand it now
two mass cancels each other
thus leaves with vertical components
Note to self: Don't be a baby and look up the solution to Real-analysis problems.Not even a single one. If you feel like giving up... then say to yourself.
12:54
hey chat
my fellow @Thorgott, I need you: suppose you have a category $\mathcal{C}$ with finite products with an object $T$ such that $(\mathcal{C}, \times, T)$ is a (strict?) monoidal category. how do I show that $T$ is a terminal object?
13:25
yeah it must be strict and it's just the projection; for any other arrow, you fill the objects with identities and use product diagram.
13:43
I am trying to find the Laplace transform of $f(x) = \sin x \cos x$ but I keep making a mistake somewhere. First, note that $\sin x \cos x = \frac{\sin (2x)}{2} = \frac{e^{2 i x} - e^{-2ix}}{4i}$, and so
\begin{align*}
(\phi f)(p) &= \int_{0}^{\infty} \sin x \cos x ~ e^{-px} dx \\
&= \frac{1}{4i} \int_{0}^{\infty} (e^{2 i x} - e^{-2ix}) e^{-px} dx \\
&= \frac{1}{4i} \left( \int_{0}^{\infty} e^{(2i-p)x)} dx - \int_{0}^{\infty} e^{-(2i+p)x} dx \right) \\
&= \frac{1}{4i} \left( \frac{1}{2i-p} e^{(2i-p)x} \bigg|_{0}^{\infty} - \frac{1}{2i+p} e^{-(2i+p)x} \bigg|_{0}^{\infty} \right) \\
But from what I am reading, the answer should be $\frac{2}{p^2 + 4}$. Where am I going wrong?
14:25
@user193319 I don't see anything wrong. Mathematica gives your answer, so that's 3 against whatever source says it is $\frac2{p^2+4}$.
However, your fourth line should be $\displaystyle\frac{1}{4i} \left( \frac{1}{2i-p} e^{(2i-p)x} \bigg|_{0}^{\infty} \color{#C00}{+} \frac{1}{2i+p} e^{-(2i+p)x} \bigg|_{0}^{\infty} \right)$
14:42
You also asked this question yesterday. I thought it looked familiar.
15:04
Anyone know where I could find a good proof of the Riemann mapping theorem? The most I've found is that it follows from the Schwarz lemma, which seems intuitively true. But all lectures / texts I've found just say "the proof is beyond the scope of this course," which, fair, but also I'd like to see one
(Is this the kind of thing that would make a good main-site question?)
15:32
Ah, nevermind, I got one.
In reference to the above linked question: suppose f is everywhere continuous on R and let A be any non-empty proper subset of R then is it true that $f(A)^o\subset f(A^o)$, where $A^o$ denotes the set of all interior points of A?
The result is false as the counter-example in the post shows.
But I don't know what's wrong in the following proof of mine:
@LucasHenrique I'm a bit iffy on what strictness means here. What you want should be that the left and right unitor are given by the canonical projections. Then what you're saying is correct.
@Slate My recollection is that Conway proves it.
15:47
Let $\displaystyle f( x) \in f( A)^{o}$. Suppose on the contrary, $\displaystyle f( x) \notin f\left( A^{o}\right)$. It follows that $\displaystyle x\notin A^{o} \Longrightarrow \forall n\ \exists x_{n} \notin S$

such that $\displaystyle x-\frac{1}{n} < x_{n} < x+\frac{1}{n} \ $...... ($\displaystyle 1^{0})$



By continuity of $ $$\displaystyle f$ at $\displaystyle x$, it follows that $\displaystyle f( x_{n})\rightarrow f( x)$. .....($\displaystyle 2^{0})$



Since $\displaystyle f( x) \in f( A)^{o} ,\ $there exists an $\displaystyle r >0$ such that $\displaystyle ( f( x) -r,f( x) +r) \in
@robjohn Oh, okay. That's good to know! I thought I was going crazy. Thanks!
What went wrong in the above proof? The claim is clearly wrong as the counterexample in post here (chat.stackexchange.com/transcript/message/59580187#59580187) shows.
$f(x_n)\in f(A)$ does not necessarily imply $x_n\in A$
@Thorgott strictness = associator and unitors are trivial (identities)
I can't think of a scenario where that ever happens
15:58
I see. $g: R\ni x\mapsto 5$ and here g(0,1)={5} and g(4) =5 but $4\notin \{5\}$.
Thanks a lot @Thorgott :)
@Thorgott ehh yeah it never happens. but there's a theorem by maclane that you can strictify a monoidal category (you get a strict monoidal category equivalent via a strong monoidal funtor to the one you're using)
@XanderHenderson Oh okay, sweet. I'll read this when I can, thank you!
it's one of those theorems that you build a structure just to work as you want and it's actually pretty neat since $V_1\otimes V_2 \otimes V_3$ is not ambiguous
16:13
@Slate NP
ok then, yeah
though, really, associators being projections suffices for this to make sense
 
1 hour later…
17:34
One of the Hatcher exercise problem is to show $\pi_1(X,x_0)\xrightarrow{j_}\pi_1(X,A,x_0)\xrightarrow{\partial}\pi_0(A,x_0)\xrightarrow{i_}\pi_0(X,x_0)$ is exact. I think I need to show $Im(j_) = ker(\partial)$ and $Im(\partial) = ker(i_)$. But what is the meaning of identity element of $\pi_0$? $\pi_0$ is not a group. I don't know how to interpret $ker(i_*)$
17:54
sodam, good question. what number exercise is it? some of the typesetting there got a little goofy.
side note, the word 'exact' and its variants appear 614 times in hatcher. this is more than one per page, on average.
2
six times in my dissertation, about .12 per page.
18:12
@leslietownes Now, is that exactly 614?
Also, does that mean that Hatcher is a long "exact" sequence?
I hear rumor that if you plug the page numbers into OEIS Hatcher will personally send you a congratulatory letter.
xander: it does.
hatcher also likes saying 'exactly' to emphasize that two conceivably different things are the same.
it's a weird verbal tic. i'm not sure this was apparent on my first readings of this book.
hence the exact science :-)
@Slate It's true! I heard it from a guy I went to graduate school with, who had a TA as an undergraduate who once dated a girl who's brother got such a letter!
i can verify that if you show up at his house enough times he'll personally send you a restraining order.
18:22
@XanderHenderson Wow, that's a great story. I can't believe I've met someone who went to graduate school with someone whose undergraduate TA dated a girl whose brother got a letter from Hatcher.
@leslietownes Exercise 4.1.8.
i want your autograph.
i'd settle for the autograph of the guy you went to graduate school with.
18:34
i'll settle for elon musk's estate.
i'm so sad that convex stuff has been co opted by ml & the like.
think of all the money you could have made with these buzzwords, if only it had been a thing a decade or two ago.
how do you make money of off buzzwords?
It involves blockchains.
yes. let me introduce you to lesliecoin.
@leslietownes you would not believe...
18:42
riiight :D
there's an indelible ledger. this is a big part of it.
everyone is saying that lesliecoin is the only thing they trust anymore, and that it is the greatest thing that has ever happened to them.
sounds like the fifth estate
@leslietownes if you ask any pure mathematician "what is [insert complicated thing they study]" they'll start with "it's just"
@LeakyNun Which is, coincidentally, how category theorists begin every sentence. :P
yeah, i was going to say. they should rename category theory "it's just"
18:55
Which do you suppose is worse? Mansplaining, or categorysplaining?
an immortal quote from an algebra class i took once was "there's no trick to this. it's just a little trick."
dunno. i find ironic categorysplaining funny. mansplaining is pretty hard to find funny.
unless it's just impossibly stupid.
I can understand why anyone would begin to explain something complicated as something simple. Many things are simpler than they appear, and principles are as simple as one gets, so a mathematician who understands a thing in his field well will know its substance well, thus the "it's just..." explanation.
I mean, how many people here remember from around middleschool when calculus and the mere concept of a derivative was considered "for smart people" or "bigbrained"?
No, small child. There are worse things than elementary calculus...
AMDG your point is well taken but it's somewhat annoying with CT, because frequently the fact that something can be phrased in category language is nowhere close to providing a proof of the truth of that thing. you often need to peek under the hood.
it's almost like saying "oh, that's just something that can be expressed in the english language."
okay? and you get grants for saying that? :)
@AMDG I disagree. I think that things actually are complicated, but years of experience makes them look simple. It is the curse of knowledge.
A plausible opinion. I will begin my counterargument by stating that this is actually simpler than it appears and that it's just...
:P
kek
at some point we have to decide what is considered "easy"
Yes, but easy how?
Easy as in intuitive to grasp or easy as in the formulas are so simple a toddler could mimic them?
easy to show
Is this still about category theory or do you mean easy to show in general?
19:10
In general, I guess.
So isn't that asking about what is considered easy to prove you might say? But if I remember correctly, finding proofs is NP-complete :)
i think most people who teach end up being quickly disabused of how easy some things are. there was a good example the other day, uniqueness of prime factorization.
well known, certainly, and easy enough to grasp. kinda hard to prove!
You might say it's a prime example.
no u
19:12
lol
people with the puns around here.
not that hard to prove, it mostly comes down to $\mathbb{Z}$ being 1-dimensional
Is uniqueness of prime factorization the claim that all prime factorizations are unique?
The fundamental theorem of arithmetic
well not hard for people used to proofs, but hard to prove in a class where people might be seeing it for the first time.
seeing it as a theorem requiring proof and not just as a fact about life.
maybe there's some slick way, i dunno. this is not a hill i will die on.
19:17
indeed
All composites are unique; given the factorization of a natural $n$, there are zero or more factors of 2. Let $k$ stand for the number of factors of 2 in the prime factorization of $n$, then $2n$ must have $k+1$ factors of 2, and all factorizations of $n\in\text{ primes}$ are necessarily unique since all primes are unique and the factorization of $p prime$ is $1p$, therefore all prime factorizations are unique?
i think the reason it's tricky, to the extent it is tricky, is that it does not follow from the ring axioms. you somehow have to leverage what makes Z Z.
of course there are a million ways to do that. i don't want to give the impression it is an unsolved problem. :D
Dang it, ran out of time to fix that, but is that a valid proof?
well, my remark was half-joking, but it also actually does capture what's going on
thorgott has the irritating habit of being right
19:26
exactly
Nevermind, I didn't think it through enough.
(No one spoil the answer for me!)
I apologize, I'll try being more wrong in the future
@leslietownes One problem is that the definition of 'prime' given to the public these days is that it only has $1$ and itself as divisors. That is the definition of 'irreducible'. The definition of prime is that if $p\mid ab$ then $p\mid a$ or $p\mid b$.
thorgott: it would help immensely, thanks.
In $\mathbb{Z}$, they are the same.
19:30
robjohn this point was made fairly well in one of the books i met early on. i forget which one.
I think I figured it out better using primorial as a basis
The public only has Z to work with.
19:44
my political platform is to offer rings other than Z for the public to work with.
@leslietownes Not the ring of truth, I hope... people won't stand for that from a politician.
@Thorgott But $\mathbb{Z}$ is zero dimensional. :(
I mean, for any reasonable notion of dimension... topological, Hausdorff, Minkowski-Bouligand, Assouad...
@XanderHenderson no it's 1-dimensional
Krull dimension
@LeakyNun That's a stupid dimension.
that's because you're looking at the wrong picture
Spec Z is 1-dimensional
Spec Z is the correct space
The only correct dimension for groups is asymptotic dimension anyway, and that gives dimension $1$ to $\Bbb Z$
What dimension does Spec Z have in the sense of covering dimension and/or inductive dimension?
i'm with leaky on this one.
the right topological dimension is the cardinality of the largest increasing chain of irreducible subsets
- some algebraic geometer
@AlessandroCodenotti i think it would have infinite covering dimension?
19:51
@XanderHenderson as a lattice, it is one-dimensional.
$\mathbb{Z}$ has covering dimension zero, assuming that you mean Lebesgue covering dimension.
As any covering of $\mathbb{Z}$ by open balls admits a refinement of non-intersecting open balls.
Spec(Z), not Z
@LeakyNun sounds suspicious
@AlessandroCodenotti you can't decrease the order by taking refinement; in fact an open cover with n open sets has order n
Fair
Dimension of nonhausdorff spaces is just wrong anyway
The definition of the inductive dimensions is even all vacuous without enough separation axioms
20:02
nonhausdorff spaces are wrongh
If $f\in C^{\infty}(M)$ and $X_p$ is a tangent vector, why in the world do we have $dfX\in T^1_1(T_pM)$
@XanderHenderson ...I see now why people leave out the proof of the mapping theorem wherever possible. One of those proofs that is high-value to have seen once, but seems low-value to rework in situ. Thanks for the reference, it was quite helpful.
Yes, all spaces are either compact Hausdorff or Polish (or both)
what's $dfX$ and what's $T_1^1$
@Alessandro hmm, not all CW complexes are polish
$T^1_1$ is space of $(1,1)$ tensors
20:05
countables are, though, right?
i'm trying to understand what $dfX$ is, I assume the prof means $(df)X$
$df$ is standard differential form
but what does $(df)X$ even mean here
he sais a connection $\nabla$ is a map $\nabla: T_pM \rightarrow T^1_1(T_pM)$ such that if $f$ smooth $\nabla(fX)=dfX+f\nablaX$
But $T^1_1(T_pM)=End(T_pM)$
they mean $X_p\otimes df\vert_p\in T_pM\otimes T_p^{\ast}M$
he hasn't discussed tensor product as of yet
which is the endomorphism $T_pM\rightarrow T_pM,\,Y_p\mapsto df\vert_p(Y_p)X_p$
then why on earth are you talking about (1,1)-tensors
he defined tensors as multilinear maps
20:11
anyway, this is a curried definition of a connection, which makes it slightly more confusing than it should be. look somewhere that defines connections as maps $TM\times TM\rightarrow TM$ and that will be clearer
20:22
@Thorgott Yes but you never encounter one that isn't
@monoidal This is horrrrrrid notation. Better to write $$\nabla_Y (fX) = (Yf)X + f\nabla_Y X$. Here $Yf = df(Y)$. Understand?
I only work with polished CW complexes.
Hi, demonic @Alessandro.
But he hasn't defined $\nabla$ on $T_pM\times T_pM$, just $T_pM$
@TedShifrin
the one you sent is the one i'm used to
I like to think of a connection on a vector bundle as giving you a vector bundle-valued $1$-form, because I like doing all of geometry with differential forms.
You can't define $\nabla$ on $T_pM$. That's just incorrect.
You need local data, not just pointwise.
I agree. He's not a mathematician. Physicist
He's writing garbage.
Sometimes lecturers are garbage.
20:29
it's annoying because there is no clarification and no brackets or anything in his writing
Well, just accept it as garbage.
On that note, I'm going to make lunch :P
@TedShifrin so late, I already finished my dinner
There are plenty of good sources to read on connections on vector bundles.
@Leaky ... There would be even one more hour difference except we just changed our clocks.
Christoffel symbols are defined as $\nabla_{e_i}e_j$ , where $e_i,e_j$ are part of a smooth local frame in the way i've been taught. Now he states: Let $\nabla$ be a connection then $\nabla(\partial_i)^{j}_k=\Gamma^{j}_{i,k}$.... how do I interpret this as in the way i've been taught... namely, in my definition, connections require two arguments
defined as "coefficients of...."
21:03
Your definition is incomplete. Care to complete it?
$\Gamma^{k}_{i,j}e_k=\nabla^{U}_{e_i}e_j$
@TedShifrin
so much of your field is decorating letters with other letters, ted. i wonder how you feel about this.
If f(x) = 1/x

f(x+h) - f(x) = h/x(x+h) ?
1/x+h - 1/x = h/x^2 + xh ?
OK, monoidal. That’s what he’s doing precisely with the coordinate frame.
@lewisdbentley You need to invest immediately in a stash of parentheses!
But I don't see how defining connection $\nabla$ on $T_pM$ is the same as defining it on $T_pM\times T_pM$
21:13
@leslietownes Hardly different from matrices …
You keep saying that wrong. Not defined on tangent space. STOP IT!
I don’t see any difference between the two.
It depends on input and output spaces.
ted: unfortunately we agree again.
my daughter yelled at a doctor today. that was kind of funny.
someone else should get it, it shouldn't always be me.
A doctor? Hers?
yes it was time for her flu shot.
the doctor asked her a series of questions clearly designed to gauge her verbal proficiency. she responded with shrugs, shaking her head, and grunts. and then "NO!" when the doctor said she wanted to look in her ears and mouth.
she can tell endless stories about her imaginary friends at home but the doctor got only one word at a time.
the doctor asked who her friends were at school. she didn't say a word. i like this, secretly. my daughter is not a snitch.
connections sorta go in one ear and out the other whenever i try to read about them
@TedShifrin That's literally what he wrote " A connection $\nabla$ is a map sending a vector $X\in T_pM$ to $\nablaX \in T^{1}_{1}(T_pM)........$
21:24
copper will know what i'm talking about. don't tell them anything.
I am not going to argue further about garbage. I have explained why it’s wrong. It is tensorial only in the vector giving the direction of differentiation (which I called $Y$). This guy doesn’t know what’s going on.
@Semiclassical differentiating sections of a vector bundle is totally natural and very central, @Semiclassic.
how should I proceed in this class then with all this garbage going on... any advice ?
Read books that are correct. What is the class and why are you taking it?
GR @TedShifrin
taking it because I like Differential Geometry
Another physicist pretending at mathematics.
Take an actual math class or go along with the physics misinformation.
21:31
i'm taking a course called Riemannian manifolds as well, taught by a mathematician, but we still haven't covered connections
Talking to the prof in his office about why he is wrong ( in a polite way) would be an option.
Riemannian geometry courses tend to do only Levi-Civita and not talk about connections in general. Complex geometers care more.
I did do that, I told him i'm unable to understand things as I know them learning DG and the way it is presented in his class. Namely, I told him the definition of a connection on a vector bundle, all he said was "that's a generalization of the directional derivative and we will get to it later"
@TedShifrin why are connections important in complex geometry?
All about vector bundles/sheaves of various sorts. I always taught more general vector bundles when I taught manifolds and diff geo. Essential for topology, too. Characteristic classes.
honestly wish you were the professor at each class i'm taking, even the ones i'm auditing, like philosophy of ethics
lol
that's where the lawyers come in. i mean, there's OK, and then there's OK OK.
21:43
Um, thanks, but I don’t know plenty!
I just have low tolerance of sloppy teaching.
i can hear squirrels romping around on my roof and this situation seems roughly analogous to that.
Who are the squirrels?
actual real-life squirrels. things that my cat takes an incredible interest in, even if she wouldn't win in a fight.
“This situation”?
people abusing notation. it's totally like squirrels romping on the roof.
21:49
Or GR prof not understanding the math right …. Not just abuse.
He is a rabid squirrel ?
i am wondering if there is any underlying understanding here.
Where?
i haven't seen squirrel froth.
i'm trying to figure out if there is a notational convention where the stated stuff makes sense.
Not if one thinks covariant derivative just depends on point values. Like calculus students love — evaluate and then differentiate. Yup.
if it feels good, then do it. that's my advice.
21:53
Much of what appears in geometry is in fact not tensorial.
This is why you’re a sleazy lawyer.
if you don't buy what i just said, i have a list of five more candidate things for you to buy.
A swamp for DeSantis?
yes we can put him in one of those. i would like to immerse him in that or something similar.
Noxious something.
let's put him in until the bubbles come up.
21:57
Let’s not get censored.
yeah, maybe better your way.
we joke about our elected representatives because they are properly subjects of public ridicule. i don't know when the next election cycle is.
i would like my local representative to lower my tax rates to zero but there seems to be no hope of that.
i would like it, though.
I sure love differentiating equalities of numbers
Mindless amusement.
 
1 hour later…
user435118
23:38
When finding the shortest distance between 2 3D lines, you find the perpendicular distance. Considering I’ve been given the equation of line A and line B in the form (xi * yj +zk) + lambda(ai + bj + ck) for both lines (mu for the second line though), I can find the vector from line A to B and my textbook says form an equation as follows: A to B DOT (a, b, c) which is the direction vector parallel to line A = 0. Do the same for line B and solve for lambda and mu.
user435118
How does this part of the method account for the initial position vectors of line A and B though? Wouldn’t the resulting A to B vector depend on the position vectors of both lines?
user435118
That should be xi + yj.

« first day (4117 days earlier)      last day (1199 days later) »