Agreed with some other comments, briefly discussing the radius of convergence would've been nice. Since n^n grows much faster than n!, this expansion only works for small z. I think it's z < 1/e if I recall correctly
"Welcome back to anow video" I love your accent and I love you. Now that my differential equations class started, I hope you keep the differential equations videos coming!
Maybe late, but here's a proof! You need to do a little work to get the expression used in the video, but gives the basic ideas. users.math.msu.edu/users/magyarp/Math880/Lagrange.pdf
I’d be curious to know how this Taylor series deals with the fact that W(z) has two branches. Does the same polynomial somehow describe both branches, or just one of them?
A really nice video - just one slight typo. For the series coefficients you define “g_n(z)” which should not be the case since you are taking the limit of the variable z (though you use x).
Omega-Tau-Phi indeed, and well deserved. One of few physicists that can do proofs, is our Papa Flammy. Remember that those physicists are the ones that were the greatest.
I was hoping to see one or two examples of using the Taylor series for W to solve a problem, or to help in a proof or a derivation? Also, does this converge rapidly enough to be useful for computing the numerical values W()?
San Samman Probably Papa Flammy's source was using this notation. That's usually a sign of advanced material. However, there is no need for operator notation here.
Hey papa flammy, not directly related to the video but I had a question. I know sometimes when you're solving your diff eqs and you have dy/dx = 2 or something, you integrate both sides with respect to x. You always say you can cancel out the dx's on the right if ur a physicist or you can introduce a proper substitution. What do you mean when you say a proper substitution?
When you use separation of variables you end up with and integral, lets call it I, that looks like I = integral f(y) dy/dx dx. Let F be an anti-derivative of f. Then F = integral f(y) dy ... dF/dy = f ... dF/dy dy/dx = f(y) dy/dx, and by the chain rule of differentiation dF/dy dy/dx is nothing but dF/dx. Therefore, I = integral dF/dx dx = F = integral f(y) dy
Akash Narayanan In a calculus book I learned from, it was proved that first order derivative, for ex. dy/dx, can be treated like a fraction, so canceling out dx is not an improper procedure. Higher order derivatives cannot be treated like fractions.
Great video as always, I've been trying to find an inverse for the anti-derivative of the Maxwell-Boltzmann boi and couldn't find any worked examples of the Lagrange inversion theorem so this helps a lot
this is wrong double checked again (the mistake was f''=-f'^2) working on this for the 2nd time i got a recursive formula for the Taylor series of the inverse of x^s*e^x but its a monster containing the sum of the n previous number in the seiris multiplied by ns however it seems like it is diverging a lot so this makes this function weird never the less it is unalitic
Lucas Depetris It seems that the constant term causes a lot of trouble. A similar equation without it is solved in wiki : en.m.wikipedia.org/wiki/Lambert_W_function ,section "Solutions of equations", example 3. However, this method cannot be applied when constant term is present, because then square root cannot be extracted. Help, Papa ! :D
Maximilian Koch I assume you want a general solution for a^x+b^x=c, since n usually denotes natural numbers and this equation is not likely to have integer solutions. It seems that this equation requires other function to solve it, since there are no x'es not in the exponent. Let's see. We want the same base for exponential functions, say, a: b^x =[a^(log_a b)]^x = a^(x*log_a b) . Let's define a parameter p: p=log_a b=ln b/ln a . Our equation becomes a^x + a^(x*p)=c -> a^x + (a^x)^p=c . Let's change a variable : a^x=y -> y+y^p=c , y>0 . So, we get a polynomial-ish equation in terms of y - not polynomial, because p may not be integer. If we get y, x is just x=ln y/ln a, ln a=/=0, a=/=1. a=1 is a trivial case. What about p=ln b/ln a ? Let's explore the cases: p=0 -> ln b=0 -> b=1, trivial. p=1 -> ln b=ln a, b=a, trivial. p=2 - quadratic equation. p=3;4 - cubic and quartic equations. Algebraic solutions are known (wiki), but they are messy, especially for quartic. We can also solve p=l/m, where l,m=1;2;3;4. For example, p=3/4 : y^(3/4) + y=c ->[y^(1/4)]^3+[y^(1/4)]^4=c y^(1/4)=t, y=t^4 ->t^3 + t^4=c If p
Maximilian Koch It seems that equation y^p +y=c can be solved by Lagrange inversion theorem. Here is solution to equation y^p - y=c : en.m.wikipedia.org/wiki/Lagrange_inversion_theorem ,Section "Example" . I guess that this equation can be turned to our form by using substitution y=(-1)^[1/(p-1)] *u, I extrapolate from article en.m.wikipedia.org/wiki/Bring_radical ,section "Normal forms", paragraph "Bring-Jerrard normal form". However, this substitution involves complex numbers in general, so... Dunno.
AncientAncestor Who likes calculating errors ? They tend to be dull, messy and tedious... Now this theorem is something else ! I suspect that proving it requires some serious knowledge of theory of functions of complex variable...
@@Hexanitrobenzene Charity Livestream? Watching Papa Flammys slow descent into madness as the hours pass of him performing increasingly more tedious but absolutely essential calculations in order to prove an absolutely essesntial result in the field of numerical integration would probably be the best thing to ever happen to the internet.
Proof of Lagrange Inversion Theorem: May y(x,b)= x + b*f(y), Near b = 0, we get the taylor expansion in b = 0: x + sum from k = 1 to infinity x^k / k! (partial^k / partial x^k) (y(x,0)) (1) y(x,b) = x - b*f(y(x,b)) partial y/partial x - 1 - b (partial f/partial y) (partial y/ partial b) = 0 means partial y/partial x (1-b*f'(y)) = 1 partial y/partial b - f(y(x,b)) - b (partial f/partial y) (partial y/ partial b) 0 means partial y/partial b (1-b*f'(y)) = f(y) Therefore, partial y/partial b = f(y) partial y / partial x (2) Now we want to show that for all n >0, (partial^n y/partial ^n b) = (partial^n-1 / partial x^n-1) (f^n (y) partial y/partial x). partial^2 y/ partial b^2 = (partial/partial b) (partial/partial b) (y) = (partial/partial b) (f(y) partial y/partial x)) = (partial/partial x) (f^2(y) partial y/partial x) By recursion, we get the said 3rd step formula. (3) In (x,0), y=x. partial y/partial x = 1 and partial^n/partial b^n (y(x,0)) = partial^n-1/partial x^n-1 (f^n(x)). By (1), we get the forth step: y = x + sum from k=1 to +infinity of b^k/k! (partial^k-1/partial x^k-1) f^k(x) (4) We have now proved the Lagrange Inversion theorem at x=0. A simple change of variable z=x+x_0 makes it in any real point.
Yay!
Prove the Lagrange Inversion Theorem
Funny seeing you here, not
Doing maths: yes papa
Proving theorem: yes papa
Lying?: no papa
don't kick me papa flammy
XD
XD
Agreed with some other comments, briefly discussing the radius of convergence would've been nice. Since n^n grows much faster than n!, this expansion only works for small z. I think it's z < 1/e if I recall correctly
"Welcome back to anow video" I love your accent and I love you. Now that my differential equations class started, I hope you keep the differential equations videos coming!
Now prove the Lagrange inversion theorem :D
Yes, would be very useful.
Yea, I really need this I'm searching this proof for days and i haven't found it
Maybe late, but here's a proof! You need to do a little work to get the expression used in the video, but gives the basic ideas.
users.math.msu.edu/users/magyarp/Math880/Lagrange.pdf
I’d be curious to know how this Taylor series deals with the fact that W(z) has two branches. Does the same polynomial somehow describe both branches, or just one of them?
Bois - T Series
Men - Pewdiepie
Legends - Flammable Maths ❤️
Two boards. Shlt is about to get serious
""It's bloody messy" 😂😂 thanks papa
But why is it called the Lambert W function? A guy named Lambert just liked the letter W?
Should pronounce it Lambert Vvvvv function, cuz that's much more deutlich, obviously.
Maybe because "L" is often used for Laplace transform and Lagrangian:)
W for weed
It's a Kamen Rider W reference
W for Wednesday my dude aaAaAAAaaAaaAAaaaAaaAaAaAAAHhhHhHHhHH
We might need to find its radius and interval of convergence. Maybe calculus stuffs, bprp can help!
ˋˋit falls a bit from the skyˋˋ
Lambert W function is the best function in the universe :V!!!!!!!
How does this work with the fact that W(z) has infinitely many branches. This definition only gives the principal branch.
Yaaay papa finally uploaded it ..
A really nice video - just one slight typo. For the series coefficients you define “g_n(z)” which should not be the case since you are taking the limit of the variable z (though you use x).
Prove the Lagrange Inversion Theorem.
Omega-Tau-Phi indeed, and well deserved. One of few physicists that can do proofs, is our Papa Flammy. Remember that those physicists are the ones that were the greatest.
So the mathematicians are the greatest physicists?
this function is a big W.
-- Labert
I was hoping to see one or two examples of using the Taylor series for W to solve a problem, or to help in a proof or a derivation?
Also, does this converge rapidly enough to be useful for computing the numerical values W()?
You seem like a chill guy... subbed
TheUnorthodoxGears
Hm, for me he looks like an arrogant smart ass, but in a funny and likeable way :D
i got confused by the D^(n -1), thanks for showing the steps!
#wewantmore ummm.. i don't know I have no criticism but I want more.
San Samman
Probably Papa Flammy's source was using this notation. That's usually a sign of advanced material. However, there is no need for operator notation here.
0:54 Mission failed. We'll get'em next time.
I come for the math, I stay cause papa is one sexy boi.
many thanks pappa
THE BOARD UNDER THE BOARD🤣🤣
MAKE MATHS GREAT AGAIN!
Hey papa flammy, not directly related to the video but I had a question. I know sometimes when you're solving your diff eqs and you have dy/dx = 2 or something, you integrate both sides with respect to x. You always say you can cancel out the dx's on the right if ur a physicist or you can introduce a proper substitution. What do you mean when you say a proper substitution?
When you use separation of variables you end up with and integral, lets call it I, that looks like
I = integral f(y) dy/dx dx.
Let F be an anti-derivative of f. Then
F = integral f(y) dy ... dF/dy = f ... dF/dy dy/dx = f(y) dy/dx,
and by the chain rule of differentiation
dF/dy dy/dx
is nothing but
dF/dx.
Therefore,
I = integral dF/dx dx = F = integral f(y) dy
Akash Narayanan
In a calculus book I learned from, it was proved that first order derivative, for ex. dy/dx, can be treated like a fraction, so canceling out dx is not an improper procedure.
Higher order derivatives cannot be treated like fractions.
Automatic captioning at 8:10 - WTF I am gonna call BND you sneaky boi
what's the radius of the convergence of this series ?
Flammy lamby!
From which lecture in the university can I learn Lagrange inversion formula?
Could you make a video on Lagrange multipliers to find functions of several variables extrema?
Beautiful :)
great video.
can someone please explain to me how he found that second derivative?
Hell yeah
Great video as always, I've been trying to find an inverse for the anti-derivative of the Maxwell-Boltzmann boi and couldn't find any worked examples of the Lagrange inversion theorem so this helps a lot
I still can't understand Taylor Functions... did you do a video about this?
Also, what is that big D at 1:05?
That's mine.
Oh okay, thanks papa!
@@NoNTr1v1aL Lol
@@misotanniold787 okay, I got the idea. Now I have to elaborate on this argument! Thank you!
I wish he would slow down more for high school students watching this.
Please try to derive Lagrange Inversion Theorem in similar way.
It is very interesting tool.
Best regards
It would really be nice if you could provide an hemdsärmeligen proof of Lagrange’s inversion theorem. ;-)
i tried it with just tayloer seiries and got y(c)+lnu(x-c) u is constent
sorry ln 1+u(x-c)
this is wrong double checked again (the mistake was f''=-f'^2) working on this for the 2nd time i got a recursive formula for the Taylor series of the inverse of x^s*e^x but its a monster containing the sum of the n previous number in the seiris multiplied by ns
however it seems like it is diverging a lot so this makes this function weird never the less it is unalitic
You have me hit mathematical climax with these series, papa.
It wpuld be great if someone sent you a featured video where the proof is explained *wink wink wonk
Hi, is it even possible to solve for x in 3^x+x^2-2=0 ?? Make a video plsss
Lucas Depetris
It seems that the constant term causes a lot of trouble. A similar equation without it is solved in wiki :
en.m.wikipedia.org/wiki/Lambert_W_function
,section "Solutions of equations", example 3. However, this method cannot be applied when constant term is present, because then square root cannot be extracted.
Help, Papa ! :D
I still want to see a proof of the Lagrange inversion theorem so that I don't have to keep using series reversions:/
Is 46 & pi the next version of that song?
@@WhattheHectogon Yeah it's gonna be on the new album
Fire Steinmeier! Flammy for President!
Challenge: Solve a^n+b^n=c making n the subject using lambert W-function :)
Maximilian Koch
I assume you want a general solution for
a^x+b^x=c,
since n usually denotes natural numbers and this equation is not likely to have integer solutions.
It seems that this equation requires other function to solve it, since there are no x'es not in the exponent.
Let's see. We want the same base for exponential functions, say, a:
b^x =[a^(log_a b)]^x = a^(x*log_a b) .
Let's define a parameter p:
p=log_a b=ln b/ln a .
Our equation becomes
a^x + a^(x*p)=c ->
a^x + (a^x)^p=c .
Let's change a variable :
a^x=y -> y+y^p=c , y>0 .
So, we get a polynomial-ish equation in terms of y - not polynomial, because p may not be integer. If we get y, x is just
x=ln y/ln a, ln a=/=0, a=/=1.
a=1 is a trivial case.
What about p=ln b/ln a ? Let's explore the cases:
p=0 -> ln b=0 -> b=1, trivial.
p=1 -> ln b=ln a, b=a, trivial.
p=2 - quadratic equation.
p=3;4 - cubic and quartic equations. Algebraic solutions are known (wiki), but they are messy, especially for quartic.
We can also solve p=l/m, where l,m=1;2;3;4. For example, p=3/4 :
y^(3/4) + y=c
->[y^(1/4)]^3+[y^(1/4)]^4=c
y^(1/4)=t, y=t^4
->t^3 + t^4=c
If p
Maximilian Koch
It seems that equation y^p +y=c can be solved by Lagrange inversion theorem. Here is solution to equation y^p - y=c :
en.m.wikipedia.org/wiki/Lagrange_inversion_theorem
,Section "Example" . I guess that this equation can be turned to our form by using substitution y=(-1)^[1/(p-1)] *u, I extrapolate from article
en.m.wikipedia.org/wiki/Bring_radical
,section "Normal forms", paragraph "Bring-Jerrard normal form".
However, this substitution involves complex numbers in general, so... Dunno.
Lagrange...
small people...
small...
Where are you studying????
Flammable Maths okk ty. You are doing a grt job, keep it up. U r doing graduation or msc???
do W(-π/2)
Give a proof of the error term in Simpsons Rule. I dare you!
AncientAncestor
Who likes calculating errors ? They tend to be dull, messy and tedious...
Now this theorem is something else ! I suspect that proving it requires some serious knowledge of theory of functions of complex variable...
@@Hexanitrobenzene Charity Livestream? Watching Papa Flammys slow descent into madness as the hours pass of him performing increasingly more tedious but absolutely essential calculations in order to prove an absolutely essesntial result in the field of numerical integration would probably be the best thing to ever happen to the internet.
Proof of Lagrange Inversion Theorem:
May y(x,b)= x + b*f(y),
Near b = 0, we get the taylor expansion in b = 0:
x + sum from k = 1 to infinity x^k / k! (partial^k / partial x^k) (y(x,0))
(1)
y(x,b) = x - b*f(y(x,b))
partial y/partial x - 1 - b (partial f/partial y) (partial y/ partial b) = 0 means partial y/partial x (1-b*f'(y)) = 1
partial y/partial b - f(y(x,b)) - b (partial f/partial y) (partial y/ partial b) 0 means partial y/partial b (1-b*f'(y)) = f(y)
Therefore, partial y/partial b = f(y) partial y / partial x
(2)
Now we want to show that for all n >0, (partial^n y/partial ^n b) = (partial^n-1 / partial x^n-1) (f^n (y) partial y/partial x).
partial^2 y/ partial b^2 = (partial/partial b) (partial/partial b) (y) = (partial/partial b) (f(y) partial y/partial x)) = (partial/partial x) (f^2(y) partial y/partial x)
By recursion, we get the said 3rd step formula.
(3)
In (x,0), y=x.
partial y/partial x = 1
and partial^n/partial b^n (y(x,0)) = partial^n-1/partial x^n-1 (f^n(x)).
By (1), we get the forth step:
y = x + sum from k=1 to +infinity of b^k/k! (partial^k-1/partial x^k-1) f^k(x)
(4)
We have now proved the Lagrange Inversion theorem at x=0. A simple change of variable z=x+x_0 makes it in any real point.
:")
So are we having an affair with the Lambert W function? I see through the lies
:D
fiRsT11!!
Please no more lambert videos 😖 too much for me
More lambert, more lambert more lambert