how to put a vector in function

hello all;

I want to write a function like below, and minimum it with "fminsearch" but I cant!

function F_m_p_cmp=fmp_cmp(cc)
mm=3
for ii=1:mm
    for t=1:24
    ex(ii,t)=cc(t)*(Uu(1,t,ii)+Uu(2,t,ii)+Uu(3,t,ii))+sum(Cc(:,t,ii))
    end
end
expp=p*ex;
eexp=exp(expp);
F_m_p_cp=(1/p)*log10(sum(ex));
end

" cc" should be a 1 by 24 vector. could you please help me?

[John D'Errico]

Note that you will never be happy with trying to solve a 24 variable problem with fminsearch. It simply is NOT designed for that many unknowns. Use a better optimizer, such as fmincon or fminunc.

Answer 1

[Star Strider]

You have to define ‘cc’ as a (1x24) vector as your initial parameter estimate vector ( x0 in the documentation ) to fminsearch.

You also have to pass Uu, Cc and p to your ‘fmp_cpm’ function. If Uu, Cc and p are already in your workspace, I would create an anonymous objective function to do that.

First, rewrite your function statement as:

function F_m_p_cmp=fmp_cmp(cc, Uu, Cc, p)

then create your objective function as:

objfcn = @(cc) fmp_cmp(cc, Uu, Cc, p);

so your call to fminsearch becomes:

cc0 = rand(1,24);    % Choose whatever (1x24) vector for ‘cc’ you want
cc_est = fminsearch(objfcn, cc0);

No promises because I can’t run your code, but this should get you started.

[Maryamm Bodaghi]

Dear Star; I don't know, ho to do that .I've attached m.files, this is a part of an algorithm, in which "delta" is a constant.

F_m_p(cmp)<= min(F_m_p(c))+delta

[Star Strider]

First, you need to change this line:

function F_m_p_cmp=fmp_cmp(cc,Uu,Cc,m,p)

to:

function F_m_p_cmp=fmp_cmp(cc,Uu,Cc,mm,p)

note the change from m to mm, and then eliminate this line in that function:

mm=1

because it will replace the value of mm in the argument list.

The objective function I wrote then becomes:

objfcn = @(cc) fmp_cmp(cc, Uu, Cc, mm, p);

The rest of my code stays the same. Make these changes, run your code, and see if fminsearch produces acceptable results. The output of fminsearch should be your optimised ‘cc’ parameters.

I do not know what you are doing in the rest of your code. I want to get fminsearch working with it, since that is what you asked.

[Maryamm Bodaghi]

No, it does not work!

I wrote this in a m.file:

Uu(:,:,1)=[2.32,2.31,4.56,2.31,2.32,6.35,6.29,4.35,5.58,5.14,5.09,4.65,4.88,7.79,3.74,3.95,9.00,8.04,7.71,5.50,8.01,8.03,7.94,7.85;
    2.70,2.70,5.24,2.69,2.70,2.70,3.70,6.54,6.25,5.82,5.77,4.68,1.81,4.77,0.72,5.84,7.88,5.74,6.41,4.20,6.70,6.73,6.74,6.85;
    3.22,3.35,6.74,4.72,4.20,3.04,3.00,3.69,6.40,6.02,8.97,7.73,5.76,8.77,1.72,4.04,6.08,7.14,6.81,4.60,7.91,7.93,7.94,8.05];

Cc(:,:,1)=[0.250002053874725,0.250001367642151,0.0650220910741544,0.250002329233973,0.250002387194205,1.32246077146061,1.43988776889062,2.55997315289463,11.5599694049485,12.8055718741147,13.9753641154536,4.26247199514375,4.26241275537540,4.17570088911312,4.35038877135854,3.95571400893718,3.79208245855539,4.26241486543702,5.54732196911996,6.48998154207310,4.40391888137625,4.31521236680666,3.06243185942612,2.68960719996417;
    0.639994942528071,0.639970075434734,0.0650262920189211,0.639997398806767,0.639997786743896,0.639998396205809,0.639998853332021,12.1600217702483,14.3600372757455,3.80586138627085,3.97563895836056,4.06245667957776,4.06253999928858,3.97573994528159,4.15050914305549,3.75561745453674,3.59187679600778,4.06227056870602,5.34734258672776,6.28996031584877,4.20390963005821,4.11523573732071,4.06240525800963,3.68962478536353;
    0.960385543435686,0.705615650592933,3.46502050556536,4.55562184969182,4.61007891624988,4.72250811222509,4.84000629429496,5.35999112726772,5.96004276944333,6.20562208217949,26.1754726280311,28.0623975732646,29.8623899416214,31.5754243915630,6.55073500723565,19.9556182433331,16.3921511440755,16.8624199376268,18.1472868131861,19.0899771735918,17.0039205660128,16.9152366395328,16.8624051939675,16.4896184971335;];

it=1;  mm=1;  
p=10^(-4);   delta=0.5;   eta=10^(-6);

eps_m_p=(log10(mm))/p;
objfcn = @(cc) fmp_cmp(cc, Uu, Cc, mm, p)
cc=fminsearch(objfcn,rand(1,24))

and below in anoteher m.fime, named "fmp_cmp".

function F_m_p_cmp=fmp_cmp(cc,Uu,Cc,mm,p)
mm=1
for ii=1:mm
    for t=1:24
    ex(ii,t)=cc(t)*(Uu(1,t,ii)+Uu(2,t,ii)+Uu(3,t,ii))+sum(Cc(:,t,ii))
    end
end
expp=p*ex;
eexp=exp(expp);
F_m_p_cp=(1/p)*log10(sum(eexp));
end

but this error occured: Error in fmp_cmp (line 2)

mm=1

Output argument "F_m_p_cmp" (and maybe others) not assigned during call to
"C:\Users\PC-Bodaghi\Desktop\fmp_cmp.m>fmp_cmp".

Error in @(cc)fmp_cmp(cc,Uu,Cc,mm,p)

Error in fminsearch (line 191)
fv(:,1) = funfcn(x,varargin{:});

Error in AA (line 15)
cc=fminsearch(objfcn,rand(1,24))

[Star Strider]

Change this line in your function:

F_m_p_cp=(1/p)*log10(sum(eexp));

to:

F_m_p_cmp=(1/p)*log10(sum(eexp));

Now run it and see if it works.

[Maryamm Bodaghi]

thanks so much,yes it works.

how can I set constraints? sum(cc)=1 && 0<=cc<=1

and also,tolerance of delta?

could you please help me?

[Star Strider]

My pleasure!

You cannot set constraints with fminsearch. You can normalise them inside your fmp_cmp function with:

cc = cc/sum(cc);

as the first statement in the function, but this is not actually constraining them. I can’t help you with delta. I have no idea what you are doing.

If you want to optimise with constraints, you will have to use one of the Optimization Toolbox solvers such as fmincon.