Lsqcurvefit on function depending on syms

Hi.

I'm using the int command to integrate a huge expression, using syms. The expression is saved in a variable "funk". In funk is a constant that I need to determine by fitting "funk" to my experimental data. For this I have tried to use lsqcurvefit, but I can't get this to work by simple putting "funk" in the expression where the function is suppose to be written. This is how it looks:

funk=huge expression with syms x

xdata=..

ydata=..

lsqcurvefit(@(x,xdata) funk, [some value] ,xdata,ydata)

How can I do this? Please help!

Answer 1

[Seth DeLand]

Hi Gustav,

You'll first need to use the matlabFunction command to convert the symbolic equation into a MATLAB function handle or file: http://www.mathworks.com/help/toolbox/symbolic/matlabfunction.html

There's an example here that should help: (see Figure 2) http://www.mathworks.com/tagteam/62669_91801v00_SymbolicGradient_Hessian_final.pdf

And another one here: http://www.mathworks.com/products/optimization/demos.html?file=/products/demos/shipping/optim/symbolic_optim_demo.html

Answer 2

[Gustav]

Ok, thanks!

Still, don't get it work. Heres an example (if I can get this to work it will solve my real problem):

xdata = [2;4;6]; %data

ydata = 3*sin([5;4;6])+6; %data

syms d t

uttryck=d*cos(t);

funk=matlabFunction(uttryck);

t=xdata;

d = lsqcurvefit(@(d,t) funk,7,t,ydata)

But I get the error message "Function value and YDATA sizes are incommensurate."

Simply, I'm trying to optimize d to fit the experimental data. Whats wrong?

[Seth DeLand]

Because funk is already a function handle (that's what matlabFunction created), you don't need to use the @(d,t) syntax when passing it to lsqcurvefit. You can just pass funk:

d = lsqcurvefit(funk,7,t,ydata)

Answer 3

[Gustav]

Nice! Works perfectly! Thank you so much!