Thread Subject: problem with quadgk

Hello everyone
I use quad functions to evaluate the integral value of my function:

>>
function y=el_en1(s,C)
Cprim=fnder(C);
ff=fnval(C, s);
ff1=fnval(Cprim, s);
Csek=fnder(Cprim);
ff2=fnval(Csek, s);
cprim_norm=sqrt(ff1(1,:).^2+ff1(2,:).^2);
kk=abs(ff'*ff2)/(abs(ff1)).^3;
yy=(kk.^2)'*cprim_norm';
y=sqrt(yy(1,:)^2+yy(2,:)^2);
>>

but every time when I coll quadgk (or any quad function)

>>
quadgk(@(s) el_en1(s, C0), 0, 1)
>>

I get this error message:
??? Error using ==> quadgk>finalInputChecks at 488
Output of the function must be the same size as the input.

Error in ==> quadgk>evalFun at 365
            finalInputChecks(x,fx);

Error in ==> quadgk>f1 at 382
        [y,too_close] = evalFun(tt);

Error in ==> quadgk>vadapt at 276
            [fx,too_close] = f(x);

Error in ==> quadgk at 215
    [q,errbnd] = vadapt(@f1,interval);


Can anyone help me? What does this mean and please any suggestions how can I fix it?

Thanks
Dijana

"Didi Cvet" <didi_cvet@yahoo.com> wrote in message
news:hcmmn6$m97$1@fred.mathworks.com...
> Hello everyone
> I use quad functions to evaluate the integral value of my function:
>
>>>
> function y=el_en1(s,C)
> Cprim=fnder(C);
> ff=fnval(C, s);
> ff1=fnval(Cprim, s);
> Csek=fnder(Cprim);
> ff2=fnval(Csek, s);
> cprim_norm=sqrt(ff1(1,:).^2+ff1(2,:).^2);
> kk=abs(ff'*ff2)/(abs(ff1)).^3;
> yy=(kk.^2)'*cprim_norm';
> y=sqrt(yy(1,:)^2+yy(2,:)^2);
>>>
>
> but every time when I coll quadgk (or any quad function)
>
>>>
> quadgk(@(s) el_en1(s, C0), 0, 1)
>>>
>
> I get this error message:
> ??? Error using ==> quadgk>finalInputChecks at 488
> Output of the function must be the same size as the input.

That's correct. From the reference page for QUADGK:

http://www.mathworks.com/access/helpdesk/help/techdoc/ref/quadgk.html

"The function y = fun(x) should accept a vector argument x and return a
vector result y. The integrand evaluated at each element of x."

[And yes, I'll ask our documentation staff to fix this.]

What happens if you call your function with a vector s and the C0 you refer
to in your anonymous function "adapter"? Does it return a vector the same
size as your input vector? If not, you need to fix your function so it
does.

--
Steve Lord
slord@mathworks.com
comp.soft-sys.matlab (CSSM) FAQ: http://matlabwiki.mathworks.com/MATLAB_FAQ

"Steven Lord" <slord@mathworks.com> wrote in message
news:hcmp1u$jcs$1@fred.mathworks.com...
 > What happens if you call your function with a vector s and the C0 you
refer
> to in your anonymous function "adapter"? Does it return a vector the same
> size as your input vector? If not, you need to fix your function so it
> does.

Yes, and we should mention that ARRAYFUN is a reasonable, if not so
efficient, way of getting it working. For example, if there are no other
errors to deal with, then this should work:

quadgk(@(t) arrayfun(@(s)el_en1(s, C0),t), 0, 1)

If performance is a problem, then I would try to vectorize el_en1.
--
Mike

"Michael Hosea" <Michael.Hosea@mathworks.com> wrote in message <hcnhjs$ajf$1@fred.mathworks.com>...
>
> "Steven Lord" <slord@mathworks.com> wrote in message
> news:hcmp1u$jcs$1@fred.mathworks.com...
> > What happens if you call your function with a vector s and the C0 you
> refer
> > to in your anonymous function "adapter"? Does it return a vector the same
> > size as your input vector? If not, you need to fix your function so it
> > does.
>
> Yes, and we should mention that ARRAYFUN is a reasonable, if not so
> efficient, way of getting it working. For example, if there are no other
> errors to deal with, then this should work:
>
> quadgk(@(t) arrayfun(@(s)el_en1(s, C0),t), 0, 1)
>
> If performance is a problem, then I would try to vectorize el_en1.
> --
> Mike
>

Thanks tor you reply
I tried as you suggest - with arrayfun, but it takes fairly much time to compute it and if I considerate that I take this integral as objective function in fmincon it works really much time so I guess I'll try to make my function to return a vector.
Thanks anyway
Didi