Serving the Quantitative Finance Community

 
User avatar
Cuchulainn
Topic Author
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Going off on a tangent (kind of) : find y

May 17th, 2017, 11:26 am

[$]x = y - \varepsilon \:  sin y[$] where [$]x = 0.8[$], [$]\varepsilon = 0.2[$]
 
User avatar
ppauper
Posts: 11729
Joined: November 15th, 2001, 1:29 pm

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 1:38 pm

maple gives y=0.9643338877

you can solve it by iteration
[$]y_{n+1}=0.8+0.2 \sin y_n[$]
when I tried that, it converges fairly quickly
 
User avatar
ppauper
Posts: 11729
Joined: November 15th, 2001, 1:29 pm

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 2:03 pm

if I write it as [$]1-\epsilon=y-\epsilon \sin y[$]   (why? it feels right, that's y) and expand y as a series in [$]\epsilon[$], the first 4 terms gives 0.9644........
[$]y=y_0+\epsilon y_1 +\epsilon^2 y_2 +\cdots[$]

y0=1
y1= sin 1 -1
etc
 
User avatar
BigAndyD
Posts: 27
Joined: July 10th, 2013, 12:32 pm

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 3:29 pm

To what precision?
 
User avatar
Paul
Posts: 6604
Joined: July 20th, 2001, 3:28 pm

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 3:40 pm

And where does it come from? Possible clues are Cuch's obsessions: Joyce, Russians, numerics, exp(5),...Iceland? But most likely the title of the thread. To get [$]y[$] appearing as a linear and a trig term suggests something involving both triangles and circles...And why have [$]x[$] and [$]\epsilon[$] and not just the numbers? These questions are so much more interesting than 0.964... I'm afraid.
 
User avatar
Traden4Alpha
Posts: 3300
Joined: September 20th, 2002, 8:30 pm

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 4:04 pm

It's a sine he's caught in a vicus of reiteration and the hypotenoose is tightening!
 
User avatar
Cuchulainn
Topic Author
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 4:24 pm

And where does it come from? Possible clues are Cuch's obsessions: Joyce, Russians, numerics, exp(5),...Iceland? But most likely the title of the thread. To get [$]y[$] appearing as a linear and a trig term suggests something involving both triangles and circles...And why have [$]x[$] and [$]\epsilon[$] and not just the numbers? These questions are so much more interesting than 0.964... I'm afraid.
Actually ...these are the things to do for relaxation :D

I have deliberately scoped the problem to work with numbers for the moment, to focus the mind.
ppauper gets an A+ for solving the (initial) problem. We can build on his solution (I see about 7 other solutions offhand). One small question is when the eccentricity [$]\varepsilon > 1.[$]
The answer is in the stars :)
As ppauper says: It is developing skill to answer the question that is posed.
Last edited by Cuchulainn on May 17th, 2017, 4:41 pm, edited 4 times in total.
 
User avatar
Cuchulainn
Topic Author
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 4:26 pm

It's a sine he's caught in a vicus of reiteration and the hypotenoose is tightening!
 You are one the right track. Get it working then right then optimised. It is a vicus problem, indeed.
Last edited by Cuchulainn on May 17th, 2017, 4:35 pm, edited 2 times in total.
 
User avatar
Cuchulainn
Topic Author
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Going off on a tangent (kind of) : find y

May 17th, 2017, 4:31 pm

To what precision?
ppauper has two solutions 1) Using Banach fixed point theorem (it is a contraction for [$]\varepsilon < 1[$]) and convergence is linear. It can be improved to second order by Aitken.
2) I don't know the answer in the asymptotic expansion case. But I am sure it is a piece of cake for many Wilmotters.
 
User avatar
ppauper
Posts: 11729
Joined: November 15th, 2001, 1:29 pm

Re: Going off on a tangent (kind of) : find y

May 18th, 2017, 10:43 am

One small question is when the eccentricity [$]\varepsilon > 1.[$]
The answer is in the stars :)
it's straightforward to do other expansions.

if you look at [$]x=y-\epsilon \sin y[$] with [$]x=1-\epsilon[$]

you can expand about [$]\epsilon=1[$], say [$]\epsilon=1+\mu[$]

and for large [$]\epsilon[$], then [$]1/\epsilon[$] is small and expand in that
 
User avatar
Cuchulainn
Topic Author
Posts: 20252
Joined: July 16th, 2004, 7:38 am
Location: 20, 000

Re: Going off on a tangent (kind of) : find y

May 18th, 2017, 11:08 am

I get the same answers as ppauper's findings. Fixed point needs ~ 6 iterations while Aitken does it in  3 for a tolerance of 1.0e-9.

When the eccentricity is 1.5 fixed point needs 36 iteration and Aitken in 6.

The fixed point iteration hangs when [$]\varepsilon = 2[$]. Aitken and Bessel's original solution give 2.29645562918859.
 
User avatar
ppauper
Posts: 11729
Joined: November 15th, 2001, 1:29 pm

Re: Going off on a tangent (kind of) : find y

May 18th, 2017, 2:03 pm

about eps=2.131615 something happens
the curve looks sort of like a cubic and it goes from having 1 rea solution to 3 real solutions
 
User avatar
Paul
Posts: 6604
Joined: July 20th, 2001, 3:28 pm

Re: Going off on a tangent (kind of) : find y

May 18th, 2017, 2:12 pm

Can someone explain why this is of any interest at all?!
 
User avatar
Traden4Alpha
Posts: 3300
Joined: September 20th, 2002, 8:30 pm

Re: Going off on a tangent (kind of) : find y

May 18th, 2017, 2:14 pm

Note that a Russian engineer would say the solution does not exist. The "y" inside the sine must have units of an angle but sin(y) will be a unit-less ratio of distances that cannot be subtracted from an angle (the lead "y").
 
User avatar
Traden4Alpha
Posts: 3300
Joined: September 20th, 2002, 8:30 pm

Re: Going off on a tangent (kind of) : find y

May 18th, 2017, 2:15 pm

Can someone explain why this is of any interest at all?!
Banach has appealch?