Find good starting candidates for Newton-Raphson knowing one of the solutions of a parametrized system of...
$begingroup$
I have a parameterized system of equations describing the crossed ladders problem.
$(x, y)$ are the $2$ horizontal distances respectively on the left/right of the junction of the ladders
$(a, b, c)$ are parameters for, respectively; the length of the first/second ladder and the distance between the ground and the junction.
Here is my system:
$$ (x + y)^2 (x^2 + c^2) - a^2 x^2 = 0$$
$$ (x + y)^2 (y^2 + c^2) - b^2 y^2 = 0$$
I've obtained the solution $(x, y) approx$ $(0.8572, 1.7461)$ for $(a, b, c) = (4, 3, 1)$.
Is there any way to find a function $f(a, b, c)$ that would give me a good starting candidate for my system?
Thank you.
real-analysis numerical-methods fixed-point-theorems newton-raphson
$endgroup$
add a comment |
$begingroup$
I have a parameterized system of equations describing the crossed ladders problem.
$(x, y)$ are the $2$ horizontal distances respectively on the left/right of the junction of the ladders
$(a, b, c)$ are parameters for, respectively; the length of the first/second ladder and the distance between the ground and the junction.
Here is my system:
$$ (x + y)^2 (x^2 + c^2) - a^2 x^2 = 0$$
$$ (x + y)^2 (y^2 + c^2) - b^2 y^2 = 0$$
I've obtained the solution $(x, y) approx$ $(0.8572, 1.7461)$ for $(a, b, c) = (4, 3, 1)$.
Is there any way to find a function $f(a, b, c)$ that would give me a good starting candidate for my system?
Thank you.
real-analysis numerical-methods fixed-point-theorems newton-raphson
$endgroup$
$begingroup$
@Moo presumably one for which Newton-Raphson then converges
$endgroup$
– Federico
Dec 4 '18 at 17:43
$begingroup$
Yes I want a candidate for which N-R converges. I'm trying to figure out wether or not it's possible to get an analytical form for f(a, b, c)
$endgroup$
– FredV
Dec 4 '18 at 19:16
$begingroup$
Must you use the given set of equations or may you use the equivalent quartic equation given in your reference? Have you tried a continuation method? Instead of solving just one problem you can consider a class of problems where the defining parameters are $t(a,b,c)$ and $t=0.1,0.2,0.3, dotsc,1$. I suspect that the solution corresponding to t=0.1 might be a good initial guess for the case of t=0.2 and so on. Of course, this could be wrong!
$endgroup$
– Carl Christian
Dec 6 '18 at 8:40
$begingroup$
Thanks for the problem ! I had a lot of fun with it. Cheers.
$endgroup$
– Claude Leibovici
Dec 9 '18 at 14:24
$begingroup$
It would be a good exercise for you to implement the solution of the quartic equation in $T$.
$endgroup$
– Claude Leibovici
Dec 10 '18 at 9:20
add a comment |
$begingroup$
I have a parameterized system of equations describing the crossed ladders problem.
$(x, y)$ are the $2$ horizontal distances respectively on the left/right of the junction of the ladders
$(a, b, c)$ are parameters for, respectively; the length of the first/second ladder and the distance between the ground and the junction.
Here is my system:
$$ (x + y)^2 (x^2 + c^2) - a^2 x^2 = 0$$
$$ (x + y)^2 (y^2 + c^2) - b^2 y^2 = 0$$
I've obtained the solution $(x, y) approx$ $(0.8572, 1.7461)$ for $(a, b, c) = (4, 3, 1)$.
Is there any way to find a function $f(a, b, c)$ that would give me a good starting candidate for my system?
Thank you.
real-analysis numerical-methods fixed-point-theorems newton-raphson
$endgroup$
I have a parameterized system of equations describing the crossed ladders problem.
$(x, y)$ are the $2$ horizontal distances respectively on the left/right of the junction of the ladders
$(a, b, c)$ are parameters for, respectively; the length of the first/second ladder and the distance between the ground and the junction.
Here is my system:
$$ (x + y)^2 (x^2 + c^2) - a^2 x^2 = 0$$
$$ (x + y)^2 (y^2 + c^2) - b^2 y^2 = 0$$
I've obtained the solution $(x, y) approx$ $(0.8572, 1.7461)$ for $(a, b, c) = (4, 3, 1)$.
Is there any way to find a function $f(a, b, c)$ that would give me a good starting candidate for my system?
Thank you.
real-analysis numerical-methods fixed-point-theorems newton-raphson
real-analysis numerical-methods fixed-point-theorems newton-raphson
edited Dec 5 '18 at 12:25
FredV
asked Dec 4 '18 at 17:32
FredVFredV
876
876
$begingroup$
@Moo presumably one for which Newton-Raphson then converges
$endgroup$
– Federico
Dec 4 '18 at 17:43
$begingroup$
Yes I want a candidate for which N-R converges. I'm trying to figure out wether or not it's possible to get an analytical form for f(a, b, c)
$endgroup$
– FredV
Dec 4 '18 at 19:16
$begingroup$
Must you use the given set of equations or may you use the equivalent quartic equation given in your reference? Have you tried a continuation method? Instead of solving just one problem you can consider a class of problems where the defining parameters are $t(a,b,c)$ and $t=0.1,0.2,0.3, dotsc,1$. I suspect that the solution corresponding to t=0.1 might be a good initial guess for the case of t=0.2 and so on. Of course, this could be wrong!
$endgroup$
– Carl Christian
Dec 6 '18 at 8:40
$begingroup$
Thanks for the problem ! I had a lot of fun with it. Cheers.
$endgroup$
– Claude Leibovici
Dec 9 '18 at 14:24
$begingroup$
It would be a good exercise for you to implement the solution of the quartic equation in $T$.
$endgroup$
– Claude Leibovici
Dec 10 '18 at 9:20
add a comment |
$begingroup$
@Moo presumably one for which Newton-Raphson then converges
$endgroup$
– Federico
Dec 4 '18 at 17:43
$begingroup$
Yes I want a candidate for which N-R converges. I'm trying to figure out wether or not it's possible to get an analytical form for f(a, b, c)
$endgroup$
– FredV
Dec 4 '18 at 19:16
$begingroup$
Must you use the given set of equations or may you use the equivalent quartic equation given in your reference? Have you tried a continuation method? Instead of solving just one problem you can consider a class of problems where the defining parameters are $t(a,b,c)$ and $t=0.1,0.2,0.3, dotsc,1$. I suspect that the solution corresponding to t=0.1 might be a good initial guess for the case of t=0.2 and so on. Of course, this could be wrong!
$endgroup$
– Carl Christian
Dec 6 '18 at 8:40
$begingroup$
Thanks for the problem ! I had a lot of fun with it. Cheers.
$endgroup$
– Claude Leibovici
Dec 9 '18 at 14:24
$begingroup$
It would be a good exercise for you to implement the solution of the quartic equation in $T$.
$endgroup$
– Claude Leibovici
Dec 10 '18 at 9:20
$begingroup$
@Moo presumably one for which Newton-Raphson then converges
$endgroup$
– Federico
Dec 4 '18 at 17:43
$begingroup$
@Moo presumably one for which Newton-Raphson then converges
$endgroup$
– Federico
Dec 4 '18 at 17:43
$begingroup$
Yes I want a candidate for which N-R converges. I'm trying to figure out wether or not it's possible to get an analytical form for f(a, b, c)
$endgroup$
– FredV
Dec 4 '18 at 19:16
$begingroup$
Yes I want a candidate for which N-R converges. I'm trying to figure out wether or not it's possible to get an analytical form for f(a, b, c)
$endgroup$
– FredV
Dec 4 '18 at 19:16
$begingroup$
Must you use the given set of equations or may you use the equivalent quartic equation given in your reference? Have you tried a continuation method? Instead of solving just one problem you can consider a class of problems where the defining parameters are $t(a,b,c)$ and $t=0.1,0.2,0.3, dotsc,1$. I suspect that the solution corresponding to t=0.1 might be a good initial guess for the case of t=0.2 and so on. Of course, this could be wrong!
$endgroup$
– Carl Christian
Dec 6 '18 at 8:40
$begingroup$
Must you use the given set of equations or may you use the equivalent quartic equation given in your reference? Have you tried a continuation method? Instead of solving just one problem you can consider a class of problems where the defining parameters are $t(a,b,c)$ and $t=0.1,0.2,0.3, dotsc,1$. I suspect that the solution corresponding to t=0.1 might be a good initial guess for the case of t=0.2 and so on. Of course, this could be wrong!
$endgroup$
– Carl Christian
Dec 6 '18 at 8:40
$begingroup$
Thanks for the problem ! I had a lot of fun with it. Cheers.
$endgroup$
– Claude Leibovici
Dec 9 '18 at 14:24
$begingroup$
Thanks for the problem ! I had a lot of fun with it. Cheers.
$endgroup$
– Claude Leibovici
Dec 9 '18 at 14:24
$begingroup$
It would be a good exercise for you to implement the solution of the quartic equation in $T$.
$endgroup$
– Claude Leibovici
Dec 10 '18 at 9:20
$begingroup$
It would be a good exercise for you to implement the solution of the quartic equation in $T$.
$endgroup$
– Claude Leibovici
Dec 10 '18 at 9:20
add a comment |
1 Answer
1
active
oldest
votes
$begingroup$
You could make at least life simpler writing $$(x + y)^2 (x^2 + c^2) = a^2 x^2tag 1$$
$$(x + y)^2 (y^2 + c^2) = b^2 y^2tag 2$$ Make the ratio
$$frac{x^2+c^2}{y^2+c^2}=frac{a^2 x^2 }{b^2 y^2 }tag 3$$ This gives two solutions
$$y_pm=pmfrac{a, c ,x}{sqrt{(b^2-a^2), x^2+b^2, c^2}}tag 4$$
Suppose that we use $y_+$ that we plug in $(1)$ to get as equation
$$left(a^2 c^4+b^2 c^4-a^2 b^2 c^2right)+x^2 left(a^4-a^2 b^2+2 b^2
c^2right)+x^4 left(b^2-a^2right)+2 a c left(c^2+x^2right) sqrt{(b^2-a^2)x^2 +b^2c^2}=0$$ and you can notice that we can solved for $x^2$ instead of $x$.
If $a >b$, this is limiting the search domain since we need $(b^2-a^2)x^2 +b^2c^2 geq 0$ that is to say (admitting that we look for $x>0$) $$0 leq x leq frac{b c}{sqrt{a^2-b^2}}$$ Next, what I should do is a Taylor expansion around $x=0$ which, assuming that $a,b,c$ are all positive to get
$$left(c^4 (a+b)^2-a^2 b^2 c^2right)+frac{ (a+b) left(a^2 b (a-b)-c^2 (a-2 b)
(a+b)right)}{b}x^2+Oleft(x^4right)$$ to get an estimate
$$x_0=sqrt{frac{a^2 b^3 c^2-b c^4 (a+b)^2}{(a+b) left(a^2 b (a-b)-c^2 (a-2 b) (a+b)right)}}$$
Applied to your case $(a, b, c) = (4, 3, 1)$, the upper bound would be $frac{3}{sqrt{7}}approx 1.134$ and $x_0=sqrt{frac{285}{434}}approx 0.810$.
Now, Newton method will generate the following iterates
$$ left(
begin{array}{cc}
n & x_n \
0 & 0.81035920 \
1 & 0.85758341 \
2 & 0.85721270 \
3 & 0.85721269
end{array}
right)$$ and, back to $(4)$, $y=1.74607506$.
Edit
We can even do better if, instead of using Taylor, we build the $[2,2]$ Padé approximant. Solving for $0$ th numerator, we get
$$x_0=sqrt{frac A {(a+b)B}}$$ where
$$A=4 b^2 c^2 left(a^4 b^3 (a-b)-a^2 b c^2 (a+b) left(a^2+a b-3 b^2right)+c^4 (a-2 b) (a+b)^3right)$$
$$B=4 a^4 b^3 (a-b)^2+c^4 (a+b)^2 left(a^3+2 a^2 b-11 a b^2+12 b^3right)+3 a^2 b^2
c^2 left(-3 a^3+6 a^2 b+a b^2-4 b^3right)$$ Applied to the example, this would give $x_0=sqrt{frac{26505}{37037}}approx 0.846$.
Newton iterates would then be
$$left(
begin{array}{cc}
n & x_n \
0 & 0.84595255 \
1 & 0.85722574 \
2 & 0.85721269
end{array}
right)$$
For the worked example, the function we look for the zero of is
$$f(x)=-7 x^4+2 left(4 sqrt{9-7 x^2}+65right) x^2+8 sqrt{9-7 x^2}-119$$ and its $[2,2]$ Padé approximant is
$$g(x)=frac{37037 x^2-26505}{35 x^2+279}$$ Just for curiosity, plot the two functions on the same graph. You would be amazed (I hope !).
We could still do better building the $[2,4]$ Padé approximant; the expression of $x_0$ would be really nasty but for the example we would have
$$h(x)=frac{74568309 x^2-54285660}{20650 x^4+85245 x^2+571428}$$ giving $x_0=sqrt{frac{18095220}{24856103} } approx 0.853$.
Update
We could make the problem simpler defining first $X=x^2$ and then $X=frac{b^2 c^2-T^2}{a^2-b^2}$. This makes the equation to be
$$-a^4 c^4-2 a^3 c^3T+ a^2left(a^2- b^2right)T^2+2 a c T^3+T^4=0tag5$$ which can be solved analytically for $T$. This quartic has two real roots since
$$Delta=-16 a^{12} c^4 (a-b)^2 (a+b)^2 left(left(a^2-b^2right)^2+27 c^4right) < 0qquad qquadforall a,b,c$$ (see here).
I suppose that the roots have opposite signs and that the one to keep is the positive root (very ugly expressions with awful radicals).
For the worked example, we should then have to solve
$$T^4+8 T^3+112 T^2-128 T-256=0$$ for which the real roots are $T=-1.06797$ and $T= 1.96375$
So, there is an explicit solution to the problem.
$endgroup$
add a comment |
Your Answer
StackExchange.ifUsing("editor", function () {
return StackExchange.using("mathjaxEditing", function () {
StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix) {
StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
});
});
}, "mathjax-editing");
StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "69"
};
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function() {
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled) {
StackExchange.using("snippets", function() {
createEditor();
});
}
else {
createEditor();
}
});
function createEditor() {
StackExchange.prepareEditor({
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader: {
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
noCode: true, onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});
}
});
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3025874%2ffind-good-starting-candidates-for-newton-raphson-knowing-one-of-the-solutions-of%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
1 Answer
1
active
oldest
votes
1 Answer
1
active
oldest
votes
active
oldest
votes
active
oldest
votes
$begingroup$
You could make at least life simpler writing $$(x + y)^2 (x^2 + c^2) = a^2 x^2tag 1$$
$$(x + y)^2 (y^2 + c^2) = b^2 y^2tag 2$$ Make the ratio
$$frac{x^2+c^2}{y^2+c^2}=frac{a^2 x^2 }{b^2 y^2 }tag 3$$ This gives two solutions
$$y_pm=pmfrac{a, c ,x}{sqrt{(b^2-a^2), x^2+b^2, c^2}}tag 4$$
Suppose that we use $y_+$ that we plug in $(1)$ to get as equation
$$left(a^2 c^4+b^2 c^4-a^2 b^2 c^2right)+x^2 left(a^4-a^2 b^2+2 b^2
c^2right)+x^4 left(b^2-a^2right)+2 a c left(c^2+x^2right) sqrt{(b^2-a^2)x^2 +b^2c^2}=0$$ and you can notice that we can solved for $x^2$ instead of $x$.
If $a >b$, this is limiting the search domain since we need $(b^2-a^2)x^2 +b^2c^2 geq 0$ that is to say (admitting that we look for $x>0$) $$0 leq x leq frac{b c}{sqrt{a^2-b^2}}$$ Next, what I should do is a Taylor expansion around $x=0$ which, assuming that $a,b,c$ are all positive to get
$$left(c^4 (a+b)^2-a^2 b^2 c^2right)+frac{ (a+b) left(a^2 b (a-b)-c^2 (a-2 b)
(a+b)right)}{b}x^2+Oleft(x^4right)$$ to get an estimate
$$x_0=sqrt{frac{a^2 b^3 c^2-b c^4 (a+b)^2}{(a+b) left(a^2 b (a-b)-c^2 (a-2 b) (a+b)right)}}$$
Applied to your case $(a, b, c) = (4, 3, 1)$, the upper bound would be $frac{3}{sqrt{7}}approx 1.134$ and $x_0=sqrt{frac{285}{434}}approx 0.810$.
Now, Newton method will generate the following iterates
$$ left(
begin{array}{cc}
n & x_n \
0 & 0.81035920 \
1 & 0.85758341 \
2 & 0.85721270 \
3 & 0.85721269
end{array}
right)$$ and, back to $(4)$, $y=1.74607506$.
Edit
We can even do better if, instead of using Taylor, we build the $[2,2]$ Padé approximant. Solving for $0$ th numerator, we get
$$x_0=sqrt{frac A {(a+b)B}}$$ where
$$A=4 b^2 c^2 left(a^4 b^3 (a-b)-a^2 b c^2 (a+b) left(a^2+a b-3 b^2right)+c^4 (a-2 b) (a+b)^3right)$$
$$B=4 a^4 b^3 (a-b)^2+c^4 (a+b)^2 left(a^3+2 a^2 b-11 a b^2+12 b^3right)+3 a^2 b^2
c^2 left(-3 a^3+6 a^2 b+a b^2-4 b^3right)$$ Applied to the example, this would give $x_0=sqrt{frac{26505}{37037}}approx 0.846$.
Newton iterates would then be
$$left(
begin{array}{cc}
n & x_n \
0 & 0.84595255 \
1 & 0.85722574 \
2 & 0.85721269
end{array}
right)$$
For the worked example, the function we look for the zero of is
$$f(x)=-7 x^4+2 left(4 sqrt{9-7 x^2}+65right) x^2+8 sqrt{9-7 x^2}-119$$ and its $[2,2]$ Padé approximant is
$$g(x)=frac{37037 x^2-26505}{35 x^2+279}$$ Just for curiosity, plot the two functions on the same graph. You would be amazed (I hope !).
We could still do better building the $[2,4]$ Padé approximant; the expression of $x_0$ would be really nasty but for the example we would have
$$h(x)=frac{74568309 x^2-54285660}{20650 x^4+85245 x^2+571428}$$ giving $x_0=sqrt{frac{18095220}{24856103} } approx 0.853$.
Update
We could make the problem simpler defining first $X=x^2$ and then $X=frac{b^2 c^2-T^2}{a^2-b^2}$. This makes the equation to be
$$-a^4 c^4-2 a^3 c^3T+ a^2left(a^2- b^2right)T^2+2 a c T^3+T^4=0tag5$$ which can be solved analytically for $T$. This quartic has two real roots since
$$Delta=-16 a^{12} c^4 (a-b)^2 (a+b)^2 left(left(a^2-b^2right)^2+27 c^4right) < 0qquad qquadforall a,b,c$$ (see here).
I suppose that the roots have opposite signs and that the one to keep is the positive root (very ugly expressions with awful radicals).
For the worked example, we should then have to solve
$$T^4+8 T^3+112 T^2-128 T-256=0$$ for which the real roots are $T=-1.06797$ and $T= 1.96375$
So, there is an explicit solution to the problem.
$endgroup$
add a comment |
$begingroup$
You could make at least life simpler writing $$(x + y)^2 (x^2 + c^2) = a^2 x^2tag 1$$
$$(x + y)^2 (y^2 + c^2) = b^2 y^2tag 2$$ Make the ratio
$$frac{x^2+c^2}{y^2+c^2}=frac{a^2 x^2 }{b^2 y^2 }tag 3$$ This gives two solutions
$$y_pm=pmfrac{a, c ,x}{sqrt{(b^2-a^2), x^2+b^2, c^2}}tag 4$$
Suppose that we use $y_+$ that we plug in $(1)$ to get as equation
$$left(a^2 c^4+b^2 c^4-a^2 b^2 c^2right)+x^2 left(a^4-a^2 b^2+2 b^2
c^2right)+x^4 left(b^2-a^2right)+2 a c left(c^2+x^2right) sqrt{(b^2-a^2)x^2 +b^2c^2}=0$$ and you can notice that we can solved for $x^2$ instead of $x$.
If $a >b$, this is limiting the search domain since we need $(b^2-a^2)x^2 +b^2c^2 geq 0$ that is to say (admitting that we look for $x>0$) $$0 leq x leq frac{b c}{sqrt{a^2-b^2}}$$ Next, what I should do is a Taylor expansion around $x=0$ which, assuming that $a,b,c$ are all positive to get
$$left(c^4 (a+b)^2-a^2 b^2 c^2right)+frac{ (a+b) left(a^2 b (a-b)-c^2 (a-2 b)
(a+b)right)}{b}x^2+Oleft(x^4right)$$ to get an estimate
$$x_0=sqrt{frac{a^2 b^3 c^2-b c^4 (a+b)^2}{(a+b) left(a^2 b (a-b)-c^2 (a-2 b) (a+b)right)}}$$
Applied to your case $(a, b, c) = (4, 3, 1)$, the upper bound would be $frac{3}{sqrt{7}}approx 1.134$ and $x_0=sqrt{frac{285}{434}}approx 0.810$.
Now, Newton method will generate the following iterates
$$ left(
begin{array}{cc}
n & x_n \
0 & 0.81035920 \
1 & 0.85758341 \
2 & 0.85721270 \
3 & 0.85721269
end{array}
right)$$ and, back to $(4)$, $y=1.74607506$.
Edit
We can even do better if, instead of using Taylor, we build the $[2,2]$ Padé approximant. Solving for $0$ th numerator, we get
$$x_0=sqrt{frac A {(a+b)B}}$$ where
$$A=4 b^2 c^2 left(a^4 b^3 (a-b)-a^2 b c^2 (a+b) left(a^2+a b-3 b^2right)+c^4 (a-2 b) (a+b)^3right)$$
$$B=4 a^4 b^3 (a-b)^2+c^4 (a+b)^2 left(a^3+2 a^2 b-11 a b^2+12 b^3right)+3 a^2 b^2
c^2 left(-3 a^3+6 a^2 b+a b^2-4 b^3right)$$ Applied to the example, this would give $x_0=sqrt{frac{26505}{37037}}approx 0.846$.
Newton iterates would then be
$$left(
begin{array}{cc}
n & x_n \
0 & 0.84595255 \
1 & 0.85722574 \
2 & 0.85721269
end{array}
right)$$
For the worked example, the function we look for the zero of is
$$f(x)=-7 x^4+2 left(4 sqrt{9-7 x^2}+65right) x^2+8 sqrt{9-7 x^2}-119$$ and its $[2,2]$ Padé approximant is
$$g(x)=frac{37037 x^2-26505}{35 x^2+279}$$ Just for curiosity, plot the two functions on the same graph. You would be amazed (I hope !).
We could still do better building the $[2,4]$ Padé approximant; the expression of $x_0$ would be really nasty but for the example we would have
$$h(x)=frac{74568309 x^2-54285660}{20650 x^4+85245 x^2+571428}$$ giving $x_0=sqrt{frac{18095220}{24856103} } approx 0.853$.
Update
We could make the problem simpler defining first $X=x^2$ and then $X=frac{b^2 c^2-T^2}{a^2-b^2}$. This makes the equation to be
$$-a^4 c^4-2 a^3 c^3T+ a^2left(a^2- b^2right)T^2+2 a c T^3+T^4=0tag5$$ which can be solved analytically for $T$. This quartic has two real roots since
$$Delta=-16 a^{12} c^4 (a-b)^2 (a+b)^2 left(left(a^2-b^2right)^2+27 c^4right) < 0qquad qquadforall a,b,c$$ (see here).
I suppose that the roots have opposite signs and that the one to keep is the positive root (very ugly expressions with awful radicals).
For the worked example, we should then have to solve
$$T^4+8 T^3+112 T^2-128 T-256=0$$ for which the real roots are $T=-1.06797$ and $T= 1.96375$
So, there is an explicit solution to the problem.
$endgroup$
add a comment |
$begingroup$
You could make at least life simpler writing $$(x + y)^2 (x^2 + c^2) = a^2 x^2tag 1$$
$$(x + y)^2 (y^2 + c^2) = b^2 y^2tag 2$$ Make the ratio
$$frac{x^2+c^2}{y^2+c^2}=frac{a^2 x^2 }{b^2 y^2 }tag 3$$ This gives two solutions
$$y_pm=pmfrac{a, c ,x}{sqrt{(b^2-a^2), x^2+b^2, c^2}}tag 4$$
Suppose that we use $y_+$ that we plug in $(1)$ to get as equation
$$left(a^2 c^4+b^2 c^4-a^2 b^2 c^2right)+x^2 left(a^4-a^2 b^2+2 b^2
c^2right)+x^4 left(b^2-a^2right)+2 a c left(c^2+x^2right) sqrt{(b^2-a^2)x^2 +b^2c^2}=0$$ and you can notice that we can solved for $x^2$ instead of $x$.
If $a >b$, this is limiting the search domain since we need $(b^2-a^2)x^2 +b^2c^2 geq 0$ that is to say (admitting that we look for $x>0$) $$0 leq x leq frac{b c}{sqrt{a^2-b^2}}$$ Next, what I should do is a Taylor expansion around $x=0$ which, assuming that $a,b,c$ are all positive to get
$$left(c^4 (a+b)^2-a^2 b^2 c^2right)+frac{ (a+b) left(a^2 b (a-b)-c^2 (a-2 b)
(a+b)right)}{b}x^2+Oleft(x^4right)$$ to get an estimate
$$x_0=sqrt{frac{a^2 b^3 c^2-b c^4 (a+b)^2}{(a+b) left(a^2 b (a-b)-c^2 (a-2 b) (a+b)right)}}$$
Applied to your case $(a, b, c) = (4, 3, 1)$, the upper bound would be $frac{3}{sqrt{7}}approx 1.134$ and $x_0=sqrt{frac{285}{434}}approx 0.810$.
Now, Newton method will generate the following iterates
$$ left(
begin{array}{cc}
n & x_n \
0 & 0.81035920 \
1 & 0.85758341 \
2 & 0.85721270 \
3 & 0.85721269
end{array}
right)$$ and, back to $(4)$, $y=1.74607506$.
Edit
We can even do better if, instead of using Taylor, we build the $[2,2]$ Padé approximant. Solving for $0$ th numerator, we get
$$x_0=sqrt{frac A {(a+b)B}}$$ where
$$A=4 b^2 c^2 left(a^4 b^3 (a-b)-a^2 b c^2 (a+b) left(a^2+a b-3 b^2right)+c^4 (a-2 b) (a+b)^3right)$$
$$B=4 a^4 b^3 (a-b)^2+c^4 (a+b)^2 left(a^3+2 a^2 b-11 a b^2+12 b^3right)+3 a^2 b^2
c^2 left(-3 a^3+6 a^2 b+a b^2-4 b^3right)$$ Applied to the example, this would give $x_0=sqrt{frac{26505}{37037}}approx 0.846$.
Newton iterates would then be
$$left(
begin{array}{cc}
n & x_n \
0 & 0.84595255 \
1 & 0.85722574 \
2 & 0.85721269
end{array}
right)$$
For the worked example, the function we look for the zero of is
$$f(x)=-7 x^4+2 left(4 sqrt{9-7 x^2}+65right) x^2+8 sqrt{9-7 x^2}-119$$ and its $[2,2]$ Padé approximant is
$$g(x)=frac{37037 x^2-26505}{35 x^2+279}$$ Just for curiosity, plot the two functions on the same graph. You would be amazed (I hope !).
We could still do better building the $[2,4]$ Padé approximant; the expression of $x_0$ would be really nasty but for the example we would have
$$h(x)=frac{74568309 x^2-54285660}{20650 x^4+85245 x^2+571428}$$ giving $x_0=sqrt{frac{18095220}{24856103} } approx 0.853$.
Update
We could make the problem simpler defining first $X=x^2$ and then $X=frac{b^2 c^2-T^2}{a^2-b^2}$. This makes the equation to be
$$-a^4 c^4-2 a^3 c^3T+ a^2left(a^2- b^2right)T^2+2 a c T^3+T^4=0tag5$$ which can be solved analytically for $T$. This quartic has two real roots since
$$Delta=-16 a^{12} c^4 (a-b)^2 (a+b)^2 left(left(a^2-b^2right)^2+27 c^4right) < 0qquad qquadforall a,b,c$$ (see here).
I suppose that the roots have opposite signs and that the one to keep is the positive root (very ugly expressions with awful radicals).
For the worked example, we should then have to solve
$$T^4+8 T^3+112 T^2-128 T-256=0$$ for which the real roots are $T=-1.06797$ and $T= 1.96375$
So, there is an explicit solution to the problem.
$endgroup$
You could make at least life simpler writing $$(x + y)^2 (x^2 + c^2) = a^2 x^2tag 1$$
$$(x + y)^2 (y^2 + c^2) = b^2 y^2tag 2$$ Make the ratio
$$frac{x^2+c^2}{y^2+c^2}=frac{a^2 x^2 }{b^2 y^2 }tag 3$$ This gives two solutions
$$y_pm=pmfrac{a, c ,x}{sqrt{(b^2-a^2), x^2+b^2, c^2}}tag 4$$
Suppose that we use $y_+$ that we plug in $(1)$ to get as equation
$$left(a^2 c^4+b^2 c^4-a^2 b^2 c^2right)+x^2 left(a^4-a^2 b^2+2 b^2
c^2right)+x^4 left(b^2-a^2right)+2 a c left(c^2+x^2right) sqrt{(b^2-a^2)x^2 +b^2c^2}=0$$ and you can notice that we can solved for $x^2$ instead of $x$.
If $a >b$, this is limiting the search domain since we need $(b^2-a^2)x^2 +b^2c^2 geq 0$ that is to say (admitting that we look for $x>0$) $$0 leq x leq frac{b c}{sqrt{a^2-b^2}}$$ Next, what I should do is a Taylor expansion around $x=0$ which, assuming that $a,b,c$ are all positive to get
$$left(c^4 (a+b)^2-a^2 b^2 c^2right)+frac{ (a+b) left(a^2 b (a-b)-c^2 (a-2 b)
(a+b)right)}{b}x^2+Oleft(x^4right)$$ to get an estimate
$$x_0=sqrt{frac{a^2 b^3 c^2-b c^4 (a+b)^2}{(a+b) left(a^2 b (a-b)-c^2 (a-2 b) (a+b)right)}}$$
Applied to your case $(a, b, c) = (4, 3, 1)$, the upper bound would be $frac{3}{sqrt{7}}approx 1.134$ and $x_0=sqrt{frac{285}{434}}approx 0.810$.
Now, Newton method will generate the following iterates
$$ left(
begin{array}{cc}
n & x_n \
0 & 0.81035920 \
1 & 0.85758341 \
2 & 0.85721270 \
3 & 0.85721269
end{array}
right)$$ and, back to $(4)$, $y=1.74607506$.
Edit
We can even do better if, instead of using Taylor, we build the $[2,2]$ Padé approximant. Solving for $0$ th numerator, we get
$$x_0=sqrt{frac A {(a+b)B}}$$ where
$$A=4 b^2 c^2 left(a^4 b^3 (a-b)-a^2 b c^2 (a+b) left(a^2+a b-3 b^2right)+c^4 (a-2 b) (a+b)^3right)$$
$$B=4 a^4 b^3 (a-b)^2+c^4 (a+b)^2 left(a^3+2 a^2 b-11 a b^2+12 b^3right)+3 a^2 b^2
c^2 left(-3 a^3+6 a^2 b+a b^2-4 b^3right)$$ Applied to the example, this would give $x_0=sqrt{frac{26505}{37037}}approx 0.846$.
Newton iterates would then be
$$left(
begin{array}{cc}
n & x_n \
0 & 0.84595255 \
1 & 0.85722574 \
2 & 0.85721269
end{array}
right)$$
For the worked example, the function we look for the zero of is
$$f(x)=-7 x^4+2 left(4 sqrt{9-7 x^2}+65right) x^2+8 sqrt{9-7 x^2}-119$$ and its $[2,2]$ Padé approximant is
$$g(x)=frac{37037 x^2-26505}{35 x^2+279}$$ Just for curiosity, plot the two functions on the same graph. You would be amazed (I hope !).
We could still do better building the $[2,4]$ Padé approximant; the expression of $x_0$ would be really nasty but for the example we would have
$$h(x)=frac{74568309 x^2-54285660}{20650 x^4+85245 x^2+571428}$$ giving $x_0=sqrt{frac{18095220}{24856103} } approx 0.853$.
Update
We could make the problem simpler defining first $X=x^2$ and then $X=frac{b^2 c^2-T^2}{a^2-b^2}$. This makes the equation to be
$$-a^4 c^4-2 a^3 c^3T+ a^2left(a^2- b^2right)T^2+2 a c T^3+T^4=0tag5$$ which can be solved analytically for $T$. This quartic has two real roots since
$$Delta=-16 a^{12} c^4 (a-b)^2 (a+b)^2 left(left(a^2-b^2right)^2+27 c^4right) < 0qquad qquadforall a,b,c$$ (see here).
I suppose that the roots have opposite signs and that the one to keep is the positive root (very ugly expressions with awful radicals).
For the worked example, we should then have to solve
$$T^4+8 T^3+112 T^2-128 T-256=0$$ for which the real roots are $T=-1.06797$ and $T= 1.96375$
So, there is an explicit solution to the problem.
edited Dec 10 '18 at 8:02
answered Dec 9 '18 at 8:27
Claude LeiboviciClaude Leibovici
120k1157132
120k1157132
add a comment |
add a comment |
Thanks for contributing an answer to Mathematics Stack Exchange!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3025874%2ffind-good-starting-candidates-for-newton-raphson-knowing-one-of-the-solutions-of%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
$begingroup$
@Moo presumably one for which Newton-Raphson then converges
$endgroup$
– Federico
Dec 4 '18 at 17:43
$begingroup$
Yes I want a candidate for which N-R converges. I'm trying to figure out wether or not it's possible to get an analytical form for f(a, b, c)
$endgroup$
– FredV
Dec 4 '18 at 19:16
$begingroup$
Must you use the given set of equations or may you use the equivalent quartic equation given in your reference? Have you tried a continuation method? Instead of solving just one problem you can consider a class of problems where the defining parameters are $t(a,b,c)$ and $t=0.1,0.2,0.3, dotsc,1$. I suspect that the solution corresponding to t=0.1 might be a good initial guess for the case of t=0.2 and so on. Of course, this could be wrong!
$endgroup$
– Carl Christian
Dec 6 '18 at 8:40
$begingroup$
Thanks for the problem ! I had a lot of fun with it. Cheers.
$endgroup$
– Claude Leibovici
Dec 9 '18 at 14:24
$begingroup$
It would be a good exercise for you to implement the solution of the quartic equation in $T$.
$endgroup$
– Claude Leibovici
Dec 10 '18 at 9:20