0
$\begingroup$
pp=ImplicitRegion[Sum[EuclideanDistance[{x,y},pt],{pt,CirclePoints[{0,0.75},
{5,0Degree},4]}]==24 &&(-1+0.04 x^2+0.8 y^2)^3==0.00032 x^2 y^3,{x,y}]
Reduce[Element[{x,y},pp],{x,y}]//LogicalExpand   

I use above method to find the intersection point value,but 2 hours past, there's no any result.
Then I plot:

a=ContourPlot[(-1+0.04 x^2+0.8 y^2)^3==0.00032 x^2 y^3,{x,-6,6},{y,-6,6}];
b=ContourPlot[Sum[EuclideanDistance[{x,y},pt],{pt,CirclePoints[{0,0.75},
{5,0Degree},4]}]==24,{x,-6,6},{y,-6,6}];
Show[a,b]

It's easy to find there are intersection points.
plot How to calculate the intersection points value?

$\endgroup$
11
  • $\begingroup$ You have two problems here: your use of inexact numbers like 0.04 and 0.8, and your use of EuclideanDistance[], which introduces an Abs[] that gives solvers trouble. $\endgroup$ Mar 29, 2018 at 17:22
  • $\begingroup$ Also, after looking into it, it seems the solutions involve the root of a degree-30 polynomial with terribly large coefficients. Do you really need an exact solution? $\endgroup$ Mar 29, 2018 at 17:28
  • $\begingroup$ I really need exact solution! Within 2 hours, system didn't give me any solvers trouble notice. $\endgroup$
    – kittygirl
    Mar 29, 2018 at 17:28
  • 1
    $\begingroup$ But it's been running for two hours, as you say. Again, consider that your asking for the exact solution might be unreasonable, when an approximate one is adequate. $\endgroup$ Mar 29, 2018 at 17:32
  • 1
    $\begingroup$ Just to make it clear: exact means a closed-form expression. Approximate solution means floats, like 0.123456. So if you want numerical values, then your problem can be easily solved. But if you actually want something exact, then you need to solve for the roots of a high-degree polynomial which takes a very long time (and might not even be possible). $\endgroup$
    – anderstood
    Mar 29, 2018 at 18:05

3 Answers 3

4
$\begingroup$

Here is a graphic method: plot the intersection points in ContourPlot, extract them, then merge close points:

f1[x_, y_] = (-1 + 0.04 x^2 + 0.8 y^2)^3 - 0.00032 x^2 y^3;
f2[x_, y_] = 
  Sum[EuclideanDistance[{x, y}, pt], {pt, 
     CirclePoints[{0, 0.75}, {5, 0 Degree}, 4]}] - 24.;
plot = ContourPlot[{f1[x, y], f2[x, y]}, {x, -6, 6}, {y, -6, 6}, 
  Contours -> {{0}}, MeshFunctions -> {f1[#, #2] - f2[#, #2] &}, 
  Mesh -> {{{0, Directive[Red, PointSize[Large]]}}}, PlotPoints -> 400]
pts = plot[[1, 1]][[First@
     Cases[plot, Point[data_] :> data, Infinity]]];
Union[pts, SameTest -> (Norm[#1 - #2] < 0.1 &)]
(* {{-4.78877, 0.46492}, {-4.2477, -0.488119}, {4.24704, -0.488501}, 
{4.78815, 0.463675}} *)

enter image description here

If you want a greater accuracy, use these points are initial conditions for FindRoot:

Table[FindRoot[{f1[x, y], f2[x, y]}, Transpose[{{x, y}, p}]], {p, pts}]
 (* {{-4.78887, 0.465097}, {-4.24761, -0.488263}, {4.24761, -0.488263}, 
  {4.78887, 0.465097}} *)
$\endgroup$
2
  • $\begingroup$ IIRC, Silvia was the first to employ MeshFunctions for stuff like this. $\endgroup$ Mar 30, 2018 at 0:02
  • $\begingroup$ I usually use one equation for the contour and the other for a mesh function: ContourPlot[f1[x, y] == 0, {x, -6, 6}, {y, -6, 6}, ContourStyle -> None, MeshFunctions -> {f2[#, #2] &}, Mesh -> {{{0, Directive[Red, PointSize[Large]]}}}]. $\endgroup$
    – Michael E2
    Mar 30, 2018 at 12:16
3
$\begingroup$

It seems to work okay with exact coefficients, over the Reals:

eqns = Sum[EuclideanDistance[{x, y}, pt], {pt, 
       CirclePoints[{0, 0.75} // Rationalize, {5, 0 Degree}, 4]}] == 
     24 && (-1 + 0.04 x^2 + 0.8 y^2)^3 == 0.00032 x^2 y^3 // 
   Rationalize;
pp = ImplicitRegion[eqns, {x, y}]
(*
ImplicitRegion[
 Sqrt[Abs[x]^2 + Abs[-(23/4) + y]^2] +
  Sqrt[Abs[-5 + x]^2 + Abs[-(3/4) + y]^2] +
  Sqrt[Abs[5 + x]^2 + Abs[-(3/4) + y]^2] +
  Sqrt[Abs[x]^2 + Abs[17/4 + y]^2] == 24 &&
   (-1 + x^2/25 + (4 y^2)/5)^3 == (x^2 y^3)/3125, {x, y}]
*)

sols = Solve[Element[{x, y}, pp], {x, y}, Reals]; // AbsoluteTiming
(*  {2.03908, Null}  *)

An approximation to sols:

{x, y} /. N@sols
(*{{-4.24761,-0.488263}, {4.24761,-0.488263}, {-4.78887,0.465097}, {4.78887,0.465097}}*)

Alternate exact solution:

gb = GroebnerBasis[Simplify[eqns, {x, y} ∈ Reals] /. Equal -> Subtract, {y, x}];

exact = Table[
   Root[Function /@ (Take[gb, j] /. {x -> #1, y -> #2}), Take[{i, 1}, j]],
   {i, 4}, {j, 2}];
exact // N
(*{{-4.78887,0.465097}, {-4.24761,-0.488263}, {4.24761,-0.488263}, {4.78887,0.465097}}*)
$\endgroup$
2
  • $\begingroup$ Hah. GroebnerBasis[] is exactly how I got the polynomials in the pictures I was trying to show the OP. $\endgroup$ Mar 30, 2018 at 0:00
  • $\begingroup$ @J.M. Ah, I didn't notice the pictures at first. I figured Solve[] must be using GroebnerBasis[]. $\endgroup$
    – Michael E2
    Mar 30, 2018 at 12:16
0
$\begingroup$

I read this thread ,I don't think I get a general method to find exact intersection.I post the script as below:

FindCrossings2D[{f_, g_}, {x_, xmin_, xmax_}, {y_, ymin_, 
ymax_}] := {x, 
y} /. (FindRoot[{f[x, y] == 0, 
   g[x, y] == 
    0}, {{x, #[[1]]}, {y, #[[2]]}}] & /@ (ContourPlot[{f[x, y] == 
     0, g[x, y] == 0}, {x, xmin, xmax}, {y, ymin, ymax}][[1, 1]]))
f[x_, y_] := (-1 + 0.04 x^2 + 0.8 y^2)^3 - 0.00032 x^2 y^3;
g[x_, y_] := 
Sum[EuclideanDistance[{x, y}, pt], {pt, 
 CirclePoints[{0, 0.75}, {5, 0 Degree}, 4]}] - 24;
pts = FindCrossings2D[{f, g}, {x, -6, 6}, {y, -6, 6}]

ContourPlot[{f[x, y] == 0, g[x, y] == 0}, {x, -6, 6}, {y, -6, 6}, 
Epilog -> {AbsolutePointSize[6], Red, Point /@ pts}]
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.