Solving non-linear equations using symbolic variables, jacobian, fsolve.

Hi there,

I am trying to solve a set of non-linear equations using Jacobian, fsolve and symbolic variables.

I am unable to figure out how to pass a jacobian computed symbolically to fsolve.

The function file is:

function [eq,J]=solve_inp(q)
    global A m n k fout t time;
    fsub=subs(fout,t,time);
    eq=matlabFunction(fsub);
    if nargout>1
        jac=jacobian(fsub);
        J=matlabFunction(jac);
    end
end

the main file is(part of):

j=1;
for time= 0:2*pi/(omega*step):2*pi/omega
    time
    if time==0
        x0=[0 5 5 5];
    else
        x0=xy(:,j-1);
    end
    options=optimoptions(@fsolve,'Algorithm','trust-region','SpecifyObjectiveGradient',true); 
    [soln,fval]=fsolve(solve_inp(q),x0,options);
    xy(:,j)=soln;
    j=j+1;
end

here 'fout' is a function of input variables and time.

[Walter Roberson]

What is q here in the context of solve_inp(q) ?

[siddartha tadepalli]

q is an array of symbolic variables to be computed.

q=sym('v',[1 2*(n+k)]);

[Walter Roberson]

Perhaps that is what should be passed in the 'vars' parameter of matlabFunction

matlabFunction(fun, 'vars', {q})

Answer 1

[Walter Roberson]

Your x0 has more than one value. That tells us that your function is a function of multiple variables. fsolve() passes all of the trial versions of the variables together as a single vector. However, matlabFunction() by default will create different variables as separate arguments.

You need to use the 'vars' option of matlabFunction to bundle all of the variables together as a single argument. Use a cell array containing a vector of the variables. For example,

vars = symvar(jac);
J = matlabFunction(jac, 'vars', {vars});

But I would recommend instead using an explicit list of variables as otherwise the ordering of initial conditions might be wrong.