MATLAB: Problem with solving equation with ‘FSOLVE’

fsolvestarting point

Hello,

I am solving a system of equations with FSOLVE. I generated starting points with

x0 = rand(2,1);

However, for some x0s, there's an error message saying

Error using trustnleqn (line 28)

Objective function is returning undefined values at initial point. FSOLVE cannot continue.

It seems that for some starting points, the fsolve cannot find solutions.

However, I want FSOLVE to immediately retry with a new random starting point rather than stopping and showing an error message. How can I do that?

Best Answer

Don't blame fsolve; the issue is you're providing solution points as initial guesses that aren't feasible for the system equations. Better solution would be to ensure the points are feasible before calling fsolve.

But, you can just try the brute-force route of putting the guess and solution calls inside a try...catch block and just keep trying...it might just work out ok.

Related Solutions

MATLAB: Using ‘fsolve’

Hi,

why not solve it symbolic:

syms f(x,y)
f(x,y) =  x.^(2-x.^(0.5))+y.^(2-y.^(0.5));
[xsol,ysol] = vpasolve(diff(f,x) + diff(f,y) == 0, [x,y], [1 3; 1 3]);
zsol = subs(f,[x,y],[xsol,ysol]);
% plot results
fsurf(f)
hold on
scatter3(double(xsol),double(ysol),double(zsol),'or','LineWidth',2,'MarkerFaceColor','r')

MATLAB: Error message from ‘fsolve’

I can’t clarify it but I can make a guess. Since fsolve uses nonlinear least-squares gradient descent algorithms, it is sensitive to the initial parameter estimates (in the documentation, x0). What probably occurred is that it encountered a local minimum and converged on it, but wasn’t at all happy with the result and threw the Warning.

The solution is to experiment with other initial parameter estimates. Eventually, you’ll find a set that leads to a solution near or at the global minimum. This is what you want.

Best Answer

Related Solutions

MATLAB: Using ‘fsolve’

MATLAB: Error message from ‘fsolve’

Related Question