« first day (3432 days earlier)      last day (1885 days later) » 
00:00 - 22:0022:00 - 00:00

00:01
@LeakyNun Want to play a blitz game maybe ? šŸ™‚
Ok wait a second
what time control
@LeakyNun ?
you decide
@LeakyNun How about 5+2 ?
00:09
@LeakyNun Gg I'm about to go now, see you
see you
 
4 hours later…
03:45
I see you
and I almost see how to prove my tetration is analytic
03:58
yeet, I think I did it wooo
0
A: Verifying tetration properties

Simply Beautiful ArtBy the mean value theorem we have $$\frac{{}^\infty a-{}^{n+1}a}{{}^\infty a-{}^na}=\frac{a^{({}^\infty a)}-a^{({}^na)}}{{}^\infty a-{}^na}=\ln(a^{a^t})\le\ln({}^\infty a)$$ for some $t\in({}^na,{}^\infty a)$. We then have $${}^\infty a-({}^\infty a-{}^{n+1}a)[\ln({}^\infty a)]^x\ge{}^\infty a...

Hopefully this doesn't have any mistakes
 
4 hours later…
07:36
there's a more general question about prime gaps that I though would be more common to read about than the twin primes question
 
2 hours later…
RPO
RPO
10:05
<a href="https://www.codecogs.com/eqnedit.php?latex=M_{(\alpha&space;,h)}=\left&space;\{&space;f(x)|x\in&space;D\:&space;and\:&space;n''\in&space;B&space;\right&space;\}" target="_blank"><img src="https://latex.codecogs.com/gif.latex?M_{(\alpha&space;,h)}=\left&space;\{&space;f(x)|x\in&space;D\:&space;and\:&space;n''\in&space;B&space;\right&space;\}" title="M_{(\alpha ,h)}=\left \{ f(x)|x\in D\: and\: n''\in B \right \}" /></a>
sorry
...
Am I allowed to do that ?
With
In fact D(a,h) is a singleton obtained from f(x,y) computation on randomly sampled records
B is bootstrap iterations quantity from 1 to 12089 to obtain D(a,h) value
M(a,h) aims to store f(x) values for each B iteration quantities
Full explanation here, thanks for help !
0
Q: Apply function on two set to create a new (+ process validation: bootstrap iterations)

RPOI am new to algebra and help would be more than welcome to tell me if the process I have built is OK, and if my attempt to apply formula on two sets to create a new one is also OK. Context I have a database containing species records (e.g. 10 different species, with 100 rows by species ; in col...

 
2 hours later…
dsm
dsm
11:46
@LeakyNun Alright, that wasn't a great approach either. I've re-done it, and feel much better about it:
$T = T_i$$^je^i\otimes e_j\Rightarrow Tv = T_i$$^jv^ie_j = T_i$$^jw^ie_j=Tw\Rightarrow T_i$$^j(v^i-w^i)e_j = 0 \Rightarrow v^i=w^i$ because $\{T_i$$^je_j\}_{i=1\rightarrow n}$ is a linearly indepedent set. Hence, $v = w$, the isometry $T$ is one-to-one, hence invertible.
using lovely Einstein (repeated-index summation) notation
 
1 hour later…
12:51
@dsm why is $T_i^j e_j$ linearly independent?
RPO
RPO
13:08
Any idea about
?
Hello! If I have a symmetrical bilinear form and a projective variety. How do I restrict that form to my variety?
RPO
RPO
Do it coude ?
I want to solve this integral $$ \int \frac{(\sin^n \theta - \sin\theta)^{1/n}\cos\theta}{\sin^{n+1}\theta}$$
Please someone tell me how how to expand something like this $x^n - x$
@RPO Can you help me?
13:32
That looks horrible. It becomes slightly less horrible if you make a trig substitution, but Iā€™m not convinced there will be a closed-form solution
dsm
dsm
@LeakyNun ahhh, my argument for linear independence has a fatal crack. OK, round 3. I'm approaching it head on to find what the inverse is:
$(Tv|Tw) = \bar{T}_i$$^jT_k$$^\ell \bar{v}^iw^k\delta_{j\ell}=\sum_j\bar{T}_i$$^jT_k$$^j \bar{v}^iw^k$, and $(v|w) = \bar{v}^iw^j\delta_{ij}=\sum_i\bar{v}^iw^i$. Equating these two, we need to have $\sum_j\bar{T}_i$$^jT_k$$^j=\delta_{ik}\Rightarrow T^\dagger T = I\Rightarrow T^{-1} = T^\dagger\Rightarrow T$ is invertible
and just assume $T$ is finite dimensional so the right inverse is automatic
14:13
@adeshmishra have you tried any algebra software like Mathematica? I'm usually afraid to solve these because almost every indefinite, ugly integral has no closed form in terms of elementary functions.
Mathematica does give an answer for the indefinite integral in terms of the Gauss hypergeometric function, but thatā€™s not really elementary @LucasHenrique
14:32
@adeshmishra Substitute $x=\sin\theta$ then $dx$ corresponds to $\cos\theta d\theta$ so you get $$\int\frac{\sqrt[n]{x^n-x}}{x^{n+1}}$$. Now substitute $u=x^{1-n}$ you get $$\frac{1}{1-n}\int \sqrt[n]{1-u} du$$ which is straight-forward to simplify
@dsm correct
the first substitution is correct, but the second seems quite wrong
Itā€™d be right if x^n were inside the root
But itā€™s not, sooo
Here are more details:
(the $x^n$ is in the root)
$$\int \frac{\sqrt[n]{x^n-x}}{x^{n+1}}=\int x^{-n-1}\sqrt[n]{x^n-x}$$
$$=\int x^{-n} \sqrt[n]{x^{-n} \cdot(x^n-x)}$$
(pulling one power of $x$ into the root)
$$=\int x^{-n} \sqrt[n]{x-x^{1-n}}$$
Now $u=x^{1-n}$ and $du= (1-n)x^{-n} dx$
should be 1- inside that second integral
14:48
Sorry where
In the second line?
Right after you said ā€œpulling one powerā€
What is wrong there?
Should be $\sqrt[n]{1-x^{1-n}}$
Oh sorry
Yes of course you are right
indeed it should be that
so now $u=x^{1-n}$
Thatā€™s important for your simplification to work
14:50
and this turns into $\sqrt[n]{1-u}$
Yes, that was a typo
but my original post already assumed to get $1-u$
Indeed you actually get $$\int x^{-n}\sqrt[n]{1-x^{1-n}}$$
Right, thatā€™s how I noticed
And now $u=x^{1-n}$
This turns nicely into $$\frac{1}{1-n}\int \sqrt[n]{1-u}\,\mathrm du$$
Yeah. So now Iā€™m wondering why Mathematica made it so complicated
Strange indeed šŸ˜„
Usually Mathematica is very good at these things
located the issue. I thought the denominator was x^n not x^(n+1)
14:55
šŸ‘
A reminder of how easy it is to turn a doable integral into an intractable one
Reminds me of that šŸ™‚
Lol. Arguably the first one is oversimplified, since itā€™s not integrable at the origin
So while the change may look all innocent, it obviously drastically changes the behavior near zero.
@MaximilianJanisch @Semiclassical Thank you.
I don't know why I always get problem in substitution of integrals, you know..
Your step when you took that $x^{n+1}$ into the numerator and took and wrote $x^1 = \sqrt[n]{x^n}$ was the key step
15:13
as far as indefinite integration is concerned, rational functions are still somewhat "nice", because we can at least more or less algorithmically tackle them
that's not as much saying that rational functions are actually nice as it is saying that indefinite integration is not nice
@Thorgott Yes
Hi, why in a Jacobian matrix the gradient vectors are horizontal and not vertical?
@Shootforthemoon Usually it is so that the Jacobian of $f$ in every point represents a linear map from $\mathbb R^m\to\mathbb R^n$ when $f$ is a map from $\mathbb R^m\to\mathbb R^n$
So for example $(\operatorname{Jac} f)(x)\cdot h$ makes sense for every $h\in\mathbb R^m$
Thanks very much!
15:29
Yeah, writing the gradient as vertical rather than horizontal is what's more odd initially (however, this pans out more nicely in other contexts)
However, I quite don't get the "physical" difference between row vectors and column vectors, if there is any
15:44
Good point... In my opinion there really isn't any... But still the multiplication of vectors is different whether you treat them as $n\times 1$ or $1\times n$ matrices
@Shootforthemoon
So in my case I am treating $h$ has a column vector
aha
Thanks again!
You are welcome
Well for example if you have two column vectors $v_1$ and $v_2$ of same dimension $n$ and you use matrix multiplication, then $v_1^\top v_2$ is a number but $v_1 v_2^\top$ is a $n\times n$ matrix and $v_1 v_2$ and $v_1^\top v_2^\top$ are not defined (with the matrix product)
$\top$ denotes transposition i.e. turning a column vector into a row vector and vice-versa
Algebraically, the difference is that they are dual to one another in a specific sense. I'm not sure whether there is any "physical" difference.
Thoughts?
0
Q: Permutations on ordered subsets of $\Bbb N$ s.t. prime elements assume equal spacing

UltradarkConsider a permutation on an ordered set, $f:\Bbb N_n\to \Bbb N_n,$ s.t. the prime elements assume equal spacing and order after the transformation, and also the composite elements are still ordered from least to greatest after the transformation. Another condition is that prime elements must occ...

dsm
dsm
@LeakyNun third times a charm :D
16:04
mh, I see @Maximilian @Thorgott
So is it just a matter of convenience to use one form or the other?
As far as the gradient is concerned, yes
16:28
can anyone state the no.of ways in which a+b+c+d = 13 where a ā‰¤ b ā‰¤ c ā‰¤ d ā‰¤ 6
the answer I got is 220, but I want to cross-verify
16:58
@Thorgott ah ok
17:16
@John I got 140
ah yes, 140 is my answer either sorry !
I got 220-80
hey max, can you do that for sum = 100 but <= 50
@MaximilianJanisch, the answer I got is, 78449
17:33
@John I got 83153
This might also be interesting:
14
Q: Integer partition with fixed number of summands but without order

LagerbaerFor a fixed $n$ and $M$, I am interested in the number of unordered non-negative integer solutions to $$\sum_{i = 1}^n a_i = M$$ Or, in other words, I am interested in the number of solutions with distinct numbers. For $n = 2$ and $M = 5$, I would consider solutions $(1,4)$ and $(4,1)$ equivalen...

Also you can use generating functions but that probably requires a computer
@MaximilianJanisch, people on MSE claim the answer as 3789, which is too small when compared to the answers what we've gotten
Where is this claimed?
I am working on a pc right now, but the python code I wrote doesn't seem to work :(
can I post links?
(By the way, my answer is almost guaranteed to be correct because it is just a Python script checking each quadruple of 4 number)
Here is my code:
7
Q: How many ordered pairs of positive integers $\{a,b,c,d\}$ are there such that $a\leq b\leq c\leq d\leq 50$ and $a+b+c+d=100$?

SkorpionHow many set of positive integers $\{a,b,c,d\}$ are there such that $ a\leq b\leq c\leq d\leq 50$ and $a+b+c+d=100$? I was thinking about using stars and bars, and it seems to work if there were only three variables: If $a\leq b\leq c\leq 50$ and $a+b+c=100$, then we can define $x=50-a,y=50-b,z...

check the comments
17:36
just a sec
sure
Oh wait
I didn't see that $a\le b\le c\le d$
I checked all possible quadruples
so my answer is way too large
wouldn't set partition consider it?
Ok so I indeed get 3789
with the additional constraint
what was the first constraint you included?
17:39
The first was $a+b+c+d=100$ and the second $a\le b\le c\le d$
for the 1st question I posed, what was the constraint you considered?
For the first question I also didn't use $a\le b\le c\le d$
so only $a+b+c+d=13$
By the way from the comments I see that reference.wolfram.com/language/ref/IntegerPartitions.html
is an implementation in Mathematica
if you have considered only $a+b+c+d =13$, the answer would be 220, wouldn't it be?
for restricted Partitions
@John Yes, the answer is 220
But I also assumed $a,b,c,d\geq1$
if you consider $a \le b \le c \le d \le 6$, the answer is 140 right?
17:43
Wait I am confused now. Give me a second
sure
@John $0\leq a$ or $1\leq a$?
ok
So:

* There are $140$ numbers $1\le a,b,c,d\le 6$ such that $a+b+c+d=13$
* There are $11$ numbers $1\le a\le b\le c\le d\le 6$ such that $a+b+c+d=13$
* There are $3780$ numbers $1\le a\le b\le c\le d\le 50$ such that $a+b+c+d=100$
* There are $83153$ numbers $1\le a,b,c,d\le 50$ such that $a+b+c+d=100$
17:47
I believe that the answer for last one is 78449, could you verify it again? I am using the same process I used for the first one
Hello all, how to prove that multiplying a 2d vector by a negative number changes its direction?
@yh05 Define direction
(after which I can give you an answer w.r.t that definition)
`Oh oh ... it's the alternative Ted.
Hey Ted :D.
Is that 'oh oh' actually an 'uh oh'?
@John Here is my code used:
from itertools import product

cnt = 0
for a, b, c, d in product(range(1, 51), repeat=4):
    if a + b + c + d == 100:
        cnt += 1

print(cnt)
17:52
Oh I thought you would have used stars or bars or something @MaximilianJanisch
Hello Ted and Ted šŸ™‚
@Shootforthemoon Columns are vectors; rows are linear maps.
Hello Maximil.
No @TedE that would be somewhat tedious. Easiest might be generating functions but I think that is still annoying
@MaximilianJanisch I don't think it's tedious, but I can't remember how to do it after so many years of not using it
@Shootforthemoon: So the gradient $\nabla f(x)$ is a vector field which corresponds to the linear map $Df(x)$.
17:54
@TedE For example here is a similar problem (but only with three numbers and already it is in my opinion annoying)
Max, thanks for the code
You're welcome @John
@TedShifrin, thanks for the lectures link (3500/10), it helped me on lagrange multipliers
Oh, sure thing. Glad it helped.
@TedE Let the A represents the tail, B represents the head of the vector. Can I define opposite direction of the vector by having the same slope as the vector, but with the new head lieing on the ray opposite to the ray AB (with right arrow on top of AB)?
17:58
What is a vector for you @yh05 ? For me it is just an n-tuple of numbers
It seems to be a directed line segment.
I'm thinking of the geometric interpretation, a directed line segment.
Why not put the tail at $B$ and the head at $A$, @yh05?
Now you can slide it around however you want. If you want the tail at $A$ instead, slide it there and make the head at $C$, appropriately placed.
@yh05 I just wanted you to give your definition of 'direction', not of 'opposite direction'
Thanks @Ted, I'll record that
@TedShifrin
18:00
LOL, TedS will work, too. :)
ahahah ok
No problem @Shootforthemoon
@TedE I don't know. How to define direction of a 2d vector?
My textbook doesn't have a definition.
I think it is easier to say when two vectors have opposite/same direction than to define "direction" in general for vectors
Is a linear map itself a vector?
ah, no ok
a vector field lol
18:04
A corollary to the Riesz representation Theorem is that every linear map from $\mathbb R^n$ to $\mathbb R$ can be represented as a vector
There are several well-known theorems in functional analysis known as the Riesz representation theorem. They are named in honor of Frigyes Riesz. This article will describe his theorem concerning the dual of a Hilbert space, which is sometimes called the Fréchetā€“Riesz theorem. For the theorems relating linear functionals to measures, see Rieszā€“Markovā€“Kakutani representation theorem. == The Hilbert space representation theorem == This theorem establishes an important connection between a Hilbert space and its continuous dual space. If the underlying field is the real numbers, the two are i...
I think it is very easy to prove in finite dimensional vector spaces actually
But a good representation for linear maps from $\mathbb R^n$ to $\mathbb R^m$ is given by matrices which also form a vector space
mh, so the gradient at a point is a vector, at all points of a function is a vector field, right?
@yh05 For non-zero vectors you can use the angle from the positive $x$-axis. Really it is better to abandon the 'angle and direction' definition of a vector imo, but anyway, can you consider $(a,b)\ne (0,0)$, and find the direction of $(a,b)$ and $(-a,-b)$. What is the direction of $(a,b)$ and for $\lambda\in\Bbb R_{\geq 0}$ what is the direction of $(\lambda a,\lambda b)$
That makes sense.
One says that two vectors $v,w$ have the same direction if there is a positive scalar $c$ so that $v = cw$.
@Shootforthemoon Yes, the gradient is a function that maps every point to a vector
18:07
Then for any vector $v$, there is a unique unit vector $\frac{1}{|v|} v$ pointing in the same direction as $v$.
Note that this map is not linear @Shootforthemoon
So many people, especially physicists, call unit vectors "direction vectors" and think of a direction as just being the same thing as a unit vector.
But the vector itself can be interpreted as a linear map
Ahh, things get complicated :p
And, as @TedShifrin has said, this vector $\nabla f(x)$ corresponds to the linear map $\mathrm df_x$
18:09
@Shootforthemoon: You might want to watch a few of my YouTube videos on such matters.
@TedE How?
I will, thanks @TedS and @Max
@Shootforthemoon No problem. In general, $(\operatorname{Jac}f)(x)$ corresponds to $\mathrm df_x:\mathbb R^m\to\mathbb R^n$ where $f:U\to\mathbb R^n$ and $U\subset\mathbb R^m$ is an open set
@yh05 Well in your case where the tail and head are free, translate the tail to the origin, and then use trig
(Also my definition of direction requires you to translate the tail to the origin)
Ok another topic, here is a cool fact I found on MSE:
$$\int_0^\infty\frac{\sinh x-x}{x^2 \sinh x}\,\mathrm dx=\ln 2$$
18:17
Also I quite don't get the last point
lol, that's a pretty significant overkill
@Thorgott What is?
This is the implicit function theorem, @Shoot.
The therefore is just saying the same thing works for $x$ and $y$ that works for $z$ if you have the appropriate nonzero partial derivative hypothesis.
This is an advanced calc-type class?
invoking the Riesz rep thm to say linear maps from R^n to R can be represented as vectors
18:20
@TedShifrin why do we ask that all the partial derivatives must be nonzero?
No, just some.
ahhh, thanks!
The gradient vector is nonzero means some nonzero component.
Oh yeah well I just wanted to use this moment to show that it also works for infinite dimensional hilbert spaces @Thor
aka, "not the zero vector"
18:20
Oh, that's explains all
But you are right I mean it is basically just definition for $\mathbb R^n$ šŸ˜…
But I like proofs like this one here anyway @Thorgott :
I will prove that the third root of $2$ is irrational.
For contradiction, assume that $\sqrt[3]{2}$ is rational. Then $\sqrt[3]{2}=\frac pq$. So $p^3=2 q^3$ i.e. $p^3=q^3+q^3$.
But this contradicts Fermat's Last Theorem 😉 QED
yeah ^^
Lol I love applications of FLT like this
the funniest part is that Fermat doesn't suffice for proving $\sqrt{2}$ is irrational
Yeah not powerful enough
But you can do $\sqrt[n]{2}$ for any $n\geq 3$
18:26
Like imagine you're a high school student or smth being asked to prove that $\sqrt[3]{2}$ is irrational and you don't know about FLT and you're like "So it suffices to prove that there are no $p, q \in \Bbb Z$ for which $p^3 = q^3 + q^3$"
I don't need to imagine I'm a high school student or smth. I would never think of FLT in a million years.
Of course, this follows from $X^3-2$ being irreducible over Q by Eisenstein
@TedShifrin truth
Hope everyone is having nice holidays
@TedShifrin I think the first thing that comes to mind when trying to prove that $\sqrt[3]{2}$ is irrational is the modularity theorem šŸ˜‰
This argument is essentially circular. Indeed, we can assume $n$ is prime (just like for FLT) and then the proof of FLT first passes from a hypothetical nontrivial solution to $a^n + b^n = c^n$ for prime $n > 2$ to a suitable "Frey curve" $y^2 = x(x-a^n)(x+b^n)$ where one has to rig certain congruential and gcd conditions on $(a,b,c)$, including that $a$, $b$, and $c$ are pairwise coprime. Yet that step applied to $(p,q,q)$ is exactly what would be the "Euclid-style" proof that $2$ is not a rational $n$th power. Hmm, another disguised version of a Euclid proof. Like the Furstenberg thing...:) — BCnrd Oct 18 '10 at 4:25
18:29
I can't imagine anybody trying to prove that a root of an integer is irrational, high school or otherwise, that doesn't end up at the usual argument
as BCnrd calls "Euclid style"
@AlessandroCodenotti Shhhh don't ruin it please šŸ™‚
Last questiona about the implicit function theorem: is the neighbourhood in which the second condition of the theorem must hold (the partial derivative different from zero) the same neighbourhood where the implicit function itself exists?
I mean, the second condition is true for a point and therefore for a sufficiently small neighbourhood, since the derivative is continuous
166
A: Awfully sophisticated proof for simple facts

Ramón BarralSeen on http://legauss.blogspot.com.es/2012/05/para-rir-ou-para-chorar-parte-13.html Theorem: $5!/2$ is even. Proof: $5!/2$ is the order of the group $A_5$. It is known that $A_5$ is a non-abelian simple group. Therefore $A_5$ is not solvable. But the Feit-Thompson Theorem asserts that every fi...

18:35
The "second part" of the Jacobian only is required to be invertible at the point (a,b) that you fixed so that $f(a,b)=0$
(as you said, it automatically follows that the second part of Jacobian is invertible in some neighborhood)
@Shoot: It need not be the same. You could have an implicit function where the derivative vanishes "by accident."
For example, the equation $F(x,y) = y^2 = 0$ defines $y=\phi(x) = 0$ everywhere, even though $\partial F/\partial y$ vanishes along the set.
Ah, thanks!
@ÍgjøgnumMeg [This one](https://mathoverflow.net/a/42513/129831) is also one of my favorites:
There are infinitely many primes because $\zeta(3)=\prod_p \frac{1}{1-p^{-3}}$ is irrational.
(Admittedly this one is also probably circular)
@TedShifrin oh, I see
I think I remember Mathein posting something about simple primary school addition being a cohomological fact
I think a non-circular argument in a similar vein can be given using an Euler product for $\pi$
@Max hilarious
I tried to teach my nephew the cohomological approach, but he didn't understand it since it's hard to motivate groups at all before they know arithmetic
Wise conclusion, alternative Ted.
I remember during my first semester, a professor tried to convince us that convolution is basically just elementary school arithmetic (I don't remember the exact analogy, sadly). He didn't succeed though.
18:42
this is all pathology
He left the room after posting that, I wonder what it means
That exit by @MikeMiller šŸ˜‚
@Thorgott A less extreme example is that the professor in first semester Analysis at my university told students to prove that $\ell_\infty$ equipped with $\|\cdot\|_\infty$ is complete. It was pretty funny when I looked at their exercise sheets
Alright I will leave you guys now.

See you
that's not too bad, but probably tough for first-semesters
see you
19:26
Is there a word for Riemannian manifolds but the charts only need to be conformal where they overlap
so if you go around a hole you might end up with a metric a constant multiple of the original
19:58
@Thorgott I mean, discrete convolution = polynomial multiplication is simple enough
Continuous convolution, not so much
20:13
@AkivaWeinberger Are you asking if given an atlas for a manifold, and the transition maps are conformal maps, what are these called?
Maybe that isn't the right wording
Basically I want the metric to be defined only up to a constant multiple
so where two charts overlap, one can be a constant multiple of the other
I think conformal has the constant depend on the point, so that's probably the wrong word
Yeah, but even polynomial multiplication isn't elementary school arithmetic. I think he was also arguing with carrying over, but I really don't remember how his argument worked.
21:16
Here also $G(x,y,z)$ locally becomes $G(x,y,Ļ†(x,y))$ because $(x_0,y_0,z_0)$ is a solution of the system, right? The condition $F_z (x_0,y_0,z_0)ā‰ 0$ does not imply that also $G_z (x_0,y_0,z_0)ā‰ 0$, but only states that in that point we can represent the solutions of the system as a function of $x$ and $y$, true?
@Thorgott yeah, itā€™s more like elementary algebra. And even there, I mean that it only requires knowledge of polynomials and the distributive property
Not elementary in the ā€œobvious to a beginnerā€ sense
I think I managed to prove my tetration to actually satisfy the tetration recurrence, initial value, and it's also analytic and completely monotone
Now pondering if I can make it unique with the condition: $$\lim_{n\to\infty}\frac{{}^\infty a -{}^{n+x}a}{{}^\infty a -{}^na}=[\ln({}^\infty a)]^x$$
@Shootforthemoon that is correct. By the implicit function Theorem we locally have $F(x,y,z)=0\implies z=\phi(x,y)$ so you just substitute this into $G(x,y,z)$
@SimplyBeautifulArt what tetration? šŸ˜„
2
Q: Verifying tetration properties

Simply Beautiful ArtIn my previous question I asked about the numerical instability and convergence of my tetration. It would seem to be the case that it converges, but suffers from catastrophic cancellation. The definition of my tetration is provided as: $${}^xa=\lim_{n\to\infty}\log_a^{\circ n}({}^\infty a-({}^\i...

$${}^xa=\lim_{n\to\infty} \log_a^{\circ n}({}^\infty a-({}^\infty a -{}^na)[\ln({}^\infty a) ]^x)$$
This is the definition, where $\log^{\circ n}$ denotes the logarithm applied n times, and the tetration on the right is the usual tetration
Ok, what led you to this tetration? šŸ˜†
21:29
Honestly my brain feels tired lol
Bruh
you look at that thing just for the fun of it?
tfw the answer is as simple as "ignore the biggest N things contributing to the error" for each N
The latter definition intuitively comes from the former limit
Ok Ill have a quick look
drop the limit, solve for ${}^{n+x}a$, log n times, and then reapply the limit
Mmm, you commented to me somewhere recently @maxi
Yes, to your brilliant computation of $\zeta(-1)$
I showed one of my friends (who doesnā€˜t know much about math) the sleight of hand that Numberphile used to ā€žproveā€œ 1+2+3+4+ā€¦=-1/12
But then I got interested in different methods to compute $\zeta(-1)$
@Simply still there?
21:40
hm
mhm*
If you want something fun you can try this:
10
A: Why does $1+2+3+\cdots = -\frac{1}{12}$?

Simply Beautiful ArtAs I recently showed in another answer, we have the wonderful pattern: $$\sum_{k=1}^n1=n\implies\int_{-1}^0x~\mathrm dx=\zeta(0)\\\sum_{k=1}^nk=\frac{n(n+1)}2\implies\int_{-1}^0\frac{x(x+1)}2~\mathrm dx=\zeta(-1)\\\sum_{k=1}^nk^2=\frac{n(n+1)(2n+1)}6\implies\int_{-1}^0\frac{x(x+1)(2x+1)}6~\mathr...

@MaximilianJanisch
Indeed very interesting
Also about this recent answer of yours:
3
A: How calculate the twelfth root of two?

Simply Beautiful ArtA much faster algorithm would be to use bisection. Example code. An even faster algorithm would be to use Newton's method. Example code. An alternative approach would be to use exponentiation and logarithms, rewriting it as $\sqrt[12]2=\exp(\ln(2)/12)$, and then computing those instead. Methods...

there is an interesting algorithm to approximate the $n$th root of an integer
I don't find that one very interesting
that you can maybe even do in your head
21:45
it's moreso just standard techniques
yes but maybe you have heard about this one:
we start with a guess $a$ for $\sqrt[12]2$
then we have $a+\epsilon=\sqrt[12]2$ for some $\epsilon$ that is hopefully small
so by taking the 12th power
that's equivalent to the Newton algorithm provided
Oh wait really
I never realized that
:O interesting
21:47
Newton's method is all about linear approximations
oh wait I am stupid... you are right
@AkivaWeinberger seems straightforward that this is the same as a homomorphism f: pi_1M -> R^+ and a metric h on the universal cover so that for any g in pi_1M, we have g^* h = f(g) h
Doubt anybody has bothered to name such a thing
Makes sense
@MikeMiller you are back, now I need to find some good applications of Fermatā€˜s Last Theorem :)
@SimplyBeautifulArt Ok so what got you into treating these many analysis problems on MSE? šŸ˜„
@MaximilianJanisch idk, I just enjoy them like most people that answer I guess
21:57
Fair point
I love that there is a pretty big community on MSE for sums/sequences/special functions/closed forms
I'm kind of curious though
@MikeMiller This is a natural structure on $S^1\times S^2$
I want to see my tetration analytically continued to other bases now
T.T but I don't think I have the computational ability at hand
21:59
What do you mean? @Simply
my tetration only works for 1 < a < e^(1/e)
when doing a^^x
You can try something such as Taylor expanding w.r.t. a to get it to other bases though
00:00 - 22:0022:00 - 00:00

« first day (3432 days earlier)      last day (1885 days later) »