Thread: Parameter estimation Nelder Mead

Originally Posted by a.mlw.walker

Code:

    //calculating physical condition parameters a and b (stored as a_global and b_global)
	if (NelderMeadSimplexMethod(n, function, x, length, &fopt, timeout, eps) == success) {
		printf("reaching to minimum ");
	} else {
		printf("timeout  ");
	}
...

static double function(int n, double x[])
{
    ...
}

Actually I'm going to go with 'Tater on this. It's not a C thing, it's a "I don't know C"-thing. For example, you are using function without your program knowing anything about it. (By that I mean it's not that C is doing something wrong, it's that you are doing something wrong because you don't understand what you are doing.)

You need to start compiling with warnings on.


Quzah.

Is there any particular reason you just keep posting dumps of numbers, instead of what we've asked for multiple times now? We don't care about your output, we want to see the code. If I have the code, I can get my own output.


Quzah.

Originally Posted by a.mlw.walker

I think you are looking at old images or something so I have reposted the first three iterations. for the first iteration sum is the same as the matlab sse value. but the next iteration is not the same, so I think arduil is right and I am not passing back the parameters correctly, but I dont think I have changed anything to stop it passing them correctly

Sounds reasonable. If you think that a_global and b_global are what parameters you're passing back, then fine, as long as you are aware that (as far as we can see) you never make use of them ever, even a little bit, while computing -- so expecting changes to those variables to do something seems foolish. Perhaps you want to initialize a and b to a_global and b_global rather than x[0] and x[1]. Or conversely, perhaps you want to change x[0] and x[1] rather than a_global and b_global.

EDIT: What I just wrote would make sense if you were actually calculating a new a and b (or x[0] and x[1], whatever). Since you are actually doing nothing of the sort, you have calculated no changed parameters, and therefore you are unable to actually change some parameters.

Please read
Cprogramming.com FAQ > Casting malloc

You never included the correct header of stdlib.h

Edit: I suggest you read up on the proper use of the keyword "extern"; you use it right before you define the function some Compilers may not like that.

This code below should NOT have the extern; and, I would NOT use the old K&R method of function parameter typing.

Code:

extern void vectorcopy(n, x, y)
int	n;
dbl	x[], y[];
{
	int	i;

	for (i=0; i<n; i++) {
		x[i] = y[i];
	}
}

Tim S.

For starters, clean up the warnings:

Code:

main.c:15: warning: missing braces around initializer
main.c:15: warning: (near initialization for `init_T[0]')
main.c: In function `function':
main.c:38: warning: implicit declaration of function `asinh'
main.c: At top level:
main.c:80: warning: return type defaults to `int'
main.c: In function `main':
main.c:81: warning: unused variable `right'
main.c:81: warning: unused variable `fall_angle'
main.c:81: warning: unused variable `fall_time'
main.c:81: warning: unused variable `guess'
main.c:81: warning: unused variable `c0'
main.c:81: warning: unused variable `xx'
main.c:82: warning: unused variable `i'
main.c:82: warning: unused variable `num_inputs'
main.c:83: warning: unused variable `y'
main.c: At top level:
main.c:61: warning: 'rotor_function' defined but not used
mat.c:15: warning: missing braces around initializer
mat.c:15: warning: (near initialization for `init_T[0]')
mat.c: In function `function':
mat.c:38: warning: implicit declaration of function `asinh'
mat.c: At top level:
mat.c:80: warning: return type defaults to `int'
mat.c: In function `main':
mat.c:81: warning: unused variable `right'
mat.c:81: warning: unused variable `fall_angle'
mat.c:81: warning: unused variable `fall_time'
mat.c:81: warning: unused variable `guess'
mat.c:81: warning: unused variable `c0'
mat.c:81: warning: unused variable `xx'
mat.c:82: warning: unused variable `i'
mat.c:82: warning: unused variable `num_inputs'
mat.c:83: warning: unused variable `y'
mat.c: At top level:
mat.c:61: warning: 'rotor_function' defined but not used
neldmead.c:11:20: values.h: No such file or directory
neldmead.c: In function `initialize':
neldmead.c:64: warning: implicit declaration of function `malloc'
neldmead.c: At top level:
neldmead.c:32: warning: 'fprint_simplex' defined but not used
neldmead.c:41: warning: 'fprint_points' defined but not used
neldmead.c:51: warning: 'fprint_generated_points' defined but not used

If you do, you'll also find that you have a main function in each of those, which you don't want if they are to somehow be linked together.


Quzah.

Originally Posted by stahta01

Are you positive you got the right mat.c; the one I downloaded did not have a main it it.

That was it. I had pasted the wrong file in on that one.


Quzah.

Most of those warnings are because of unused but declared functions - thats because i have been stripping back the code to try and find the problem, however i didnt remove the global variable names - shouldnt affect the running of it. if i put all the code back together most of them go apart from:

main.c:16: warning: missing braces around initializer
nm_t0\main.c:16: warning: (near initialization for 'init_T[0]')
:108: warning: return type defaults to 'int'
In function 'main':
warning: 'return' with no value, in function returning non-void

I think that there is definately not a main in each of the files, I understand that there can be only one main. I didnt write the function Nelder Mead, it comes from available source code. It came with an example of its use. I first used it to calcualate two parametrs, but expanding it to three and it doesnt converge......
I'll see about that external but ut was in the original code.
I added in the stdlib.h file but the final result is still incorrect. I wondered whether a tolerance somewhere is setting to to be able to stop early?
By the way, i have attached the fit that C gives and the fit that matlab gives - matlab on the left. For some reason matlab finds a far better solution...