Prev: Kalman Estimator Structure
Next: hydraulic valve modelling
From: Piter_ on 21 Apr 2010 10:51
Hi all.
I try to estimate parameters by fitting data to model.
I have this functions:
_______________________________________________
function [sse param]=cox_model_fit(t,dy,param,conc0)
dyy=cox_cyt_dis_dt(t,param,start);
sse=norm(dyy-dy)

_______________________________________________
function [dy]=cox_cyt_dis_dt(t,param,start)
very long one for reaction simulation

_______________________________________________
t - time
dy - concentration changes.
conc0 - initial concentrations.
param0 - initial guess
for estimation of parameters run command like this:

global t,dy, conc0
param = fminsearch(@cox_model_fit,param0,options)

how can I do it without use of global variables?

Thanks.
Petro




From: Alistair Templeton on 21 Apr 2010 12:01
> _________________________________________
> function [sse param]=cox_model_fit(t,dy,param,conc0)
> dyy=cox_cyt_dis_dt(t,param,start);
> sse=norm(dyy-dy)
>
> _______________________________________________
> function [dy]=cox_cyt_dis_dt(t,param,start)
> very long one for reaction simulation
>

> global t,dy, conc0
> param = fminsearch(@cox_model_fit,param0,options)
>
> how can I do it without use of global variables?
>

I'm not familiar with fminsearch, but it seems like any variables that cox_cyt_dis_dt changes should be returnable?
From: John D'Errico on 21 Apr 2010 12:17
Piter_ <x.piter(a)gmail.com> wrote in message <2de9d013-ebd1-45db-acf4-15bac9e919da(a)z3g2000yqz.googlegroups.com>...
> Hi all.
> I try to estimate parameters by fitting data to model.
> I have this functions:
> _______________________________________________
> function [sse param]=cox_model_fit(t,dy,param,conc0)
> dyy=cox_cyt_dis_dt(t,param,start);
> sse=norm(dyy-dy)
>
> _______________________________________________
> function [dy]=cox_cyt_dis_dt(t,param,start)
> very long one for reaction simulation
>
> _______________________________________________
> t - time
> dy - concentration changes.
> conc0 - initial concentrations.
> param0 - initial guess
> for estimation of parameters run command like this:
>
> global t,dy, conc0
> param = fminsearch(@cox_model_fit,param0,options)
>
> how can I do it without use of global variables?

By passing them in as arguments, using an anonymous function.
Change your functions like this...

_______________________________________________

function [sse param]=cox_model_fit(t,dy,param,conc0,start)
dyy=cox_cyt_dis_dt(t,param,start);
sse=norm(dyy-dy);
_______________________________________________

function [dy]=cox_cyt_dis_dt(t,param,start)
% very long one for reaction simulation

_______________________________________________

Finally, here is your call to fminsearch...

param = fminsearch(@(param) cox_model_fit(t,dy,param,conc0,start),param0,options)

HTH,
John
 | 
Pages: 1
Prev: Kalman Estimator Structure
Next: hydraulic valve modelling