matlabFunction is too slow

Hi all,

I am using matlabFunction to create a .m file to define the function of a vector and its Jacobian:

matlabFunction(fvec, J, 'file', filename, 'vars', {x, phi_old});

When the dimension of the vector = 50, the time cost is reasonable; but when setting the dimension to 500, it becomes extremely slow. Would you please give advice for this problem?

Thank you.

Answer 1

[Alan Weiss]

I have no general advice. Perhaps if we saw a small snippet of your code we could help.

It is possible that your Jacobian is sparse, meaning it has a lot of zeros. In that case, it might be possible to avoid using matlabFunction at all, and simply use the JacobPattern option to have sparse finite differencing instead of time-consuming analytic calculations. Or perhaps there could be a hybrid sort of solution, where you explicitly calculate analytically (with matlabFunction) only the nonzero terms in the Jacobian. But this would require you to write a custom program.

Or it could be something as simple as declaring your symbolic variables to be real, which sometimes (but not often) saves time in matlabFunction.

Sorry I have nothing so helpful. There is simply a lot of calculation to be done, and it uses a lot of memory and time.

Alan Weiss

MATLAB mathematical toolbox documentation

[Brad Carman]

What is the "JacobPattern option"?