problem with fminsearch and parameters estimation

    function [to,yo]=magda_ewa

clc
clear all
close all

global Vg Vi Rinf Cinf Gg gamma lambda alfa C1 a1 Rm Ki Kg Tinf UFR m Gd time meang meani   

UFR=3/240; %0.004166667;                %[L/min]

Gd=5.615292712;

Rinf=60/1000;                           %[L/min]

Vinf=0.5*m*100/(33*1000);               %[L] 

Cinf=33/180.16*1000*1000/100;           %[mmol/L]

Tinf=Vinf/Rinf;                         %[min]
Go=5.62;

Io=4.3;
Gg=0.1751;                      %[mmol/min]

lambda=0.00993;                 %[L/min]

alfa=0.01855;                   %[L/min]

C1=23.9395;                     %[mmol/L]

a1=0.228339456;                  %[mmol/L]

Rm=968.8786;                    %[mU/min]

m=78;                           %[kg]

Vg=11.4587;                     %[L]

Vi=3.1201;                      %[L]

time=[0 10 20 30 40 60];

meang=[5.62 19.12 14.76 12.07 10.57 7.95];

meani=[4.3 33.98 24.64 19.03 15.38 14.84];

[to yo]=ode45('model_magda',0:60,[Go Io]);

[to yo(:,1) yo(:,2)];

--------------------------------------

function F=model_magda(t,y)

F=zeros(2,1);

global Vg Vi Rinf Cinf Gg gamma lambda alfa C1 a1 Rm Ki Kg Tinf UFR Gd Vglucose Vinsulin 

if t<Tinf 

      Ginf=Rinf*Cinf;
  else
      Ginf=0;
  end;

for t=1:61

Vglucose(t)=Vg-UFR*(t-1);

Vinsulin(t)=Vi-UFR*(t-1);

F(1)=(Gg-gamma*y(1)*y(2)-lambda*y(1)+Ginf-Kg*(y(1)-Gd)+y(1)*UFR)/Vglucose(t);

F(2)=(Rm/(1+exp((C1-y(1))/a1))-alfa*y(2)-Ki*y(2))/Vinsulin(t);

end

----------------------------

function f=minimum(par)

global  gamma %lambda alfa C1 a1 Rm Vg Vi Gg time meang meani 

gamma=par(1); 

[to,yo]=magda_ewa;

f=0;

for i=1:6

c(i)=(abs(meani(i) - yo((time(i)+1),2))/meani(i))+(abs(meang(i) - 
yo((time(i))+1,1))/meang(i));

f=f+c(i);

i=i+1;

end

end 

---------------------

clear all 

close all

[wynik,c]=fminsearch(@minimum,0.003);

gamma=wynik(1)

This is my program and when i try to run it i get this comment :

???  In an assignment  A(I) = B, the number of elements in B and
 I must be the same.

Error in ==> model_magda at 24

F(1)=(Gg-gamma*y(1)*y(2)-lambda*y(1)+Ginf-Kg*(y(1)-Gd)+y(1)*UFR)/Vglucose(t);

Error in ==> odearguments at 98

f0 = feval(ode,t0,y0,args{:});   % ODE15I sets args{1} to yp0.

Error in ==> ode45 at 172

[neq, tspan, ntspan, next, t0, tfinal, tdir, y0, f0, odeArgs, odeFcn, ...

Error in ==> magda_ewa at 211

[to yo]=ode45('model_magda',0:60,[Go Io]);

Error in ==> minimum at 15

[to,yo]=magda_ewa;

Error in ==> fminsearch at 191

fv(:,1) = funfcn(x,varargin{:});

Error in ==> optymalizacja at 3

[wynik,c]=fminsearch(@minimum,0.003);

before used fminsearch everything was ok I didn't have problems with number of elements in matrix....Please help me because i don't known wher the mistake is

[Matt J]

Please use the "{} Code" formatting tool, so that we can distinguish your code/errors from text with less strain.

Answer 1

[Matt J]

Error in model_magda at 24

Some of your variables are empty/undefined in model_magda:

    K>> whos Kg Ki Tinf gamma
      Name       Size            Bytes  Class     Attributes

      Kg         0x0                 0  double    global    
      Ki         0x0                 0  double    global    
      Tinf       0x0                 0  double    global    
      gamma      0x0                 0  double    global

Using global variables is probably a bad idea. See here for better options

http://www.mathworks.com/help/optim/ug/passing-extra-parameters.html

Using "clear all" while using globals is probably making things worse.

[Ewa Kozub]

thanks for your answer :) i'm quite new in Matlab and could you tell me how you check those parameters (Kg, Ki Tinf, gamma)?. And when i want to use function in this way F = model_magda(Vg Vi Rinf Cinf Gg gamma lambda alfa C1 a1 Rm Ki Kg Tinf UFR Gd Vglucose Vinsulin ) I get information :too many input arguments

[Matt J]

could you tell me how you check those parameters (Kg, Ki Tinf, gamma)?

By using the dbstop command. In particular

 >> dbstop if error

will halt program execution right at the point in the code where an error occurs, so you can inspect the state of the workspace variables at that moment.

I get information :too many input arguments

You have to rewrite

 function F=model_magda(t,y)

so that it knows to expect more than just t and y as inputs.