Calling function in fminsearch

Question

Yokuna 2022-10-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1832828-calling-function-in-fminsearch

评论： Steven Lord 2022-10-22

采纳的回答： Steven Lord

在 MATLAB Online 中打开

I want to call a function lamda to update initial value in fminsearch.

The initial value of lambda(1) =-1.

I want to update in lambda as lambda(i+1)=lambda(i)-0.01*(lambda(end)).

This computation of lambda should continue until lambda(end) < 0.01. I am not able to call function in fminsearch as FMINSEARCH accepts inputs only of data type double.

close all
clear all
clc
[optimal_lambda,cost] = fminsearch(@cost_lambda, @lamda)
a = optimal_lambda(1)
function cost = cost_lambda(lambda)
   part_a = @(t,y)[-0.02*y(1); (y(2)/0.8)*(1/0.8)*exp(0.01*t)];
   [t,y]=ode45(part_a, [0 10], [0.5,-1]');
   cost = (1-(1/0.01)*(1-exp(-log((-1+(y(1)/0.8)^2)/(0.015*lambda(1)*exp(0.01*t)))))*(1-(y(1)/0.8))^2);
   
function lambda = lamda(lambda)
    lambda(1)=-1;
    i=1;
        for lambda(end)<0.01
Invalid expression. When calling a function or indexing a variable, use parentheses. Otherwise, check for mismatched delimiters.

Error in connector.internal.fevalMatlab

Error in connector.internal.fevalJSON
    lambda(i+1)=lambda(i)-0.01*lambda(end)
    i=i++
        end
end
end

3 个评论
显示 1更早的评论隐藏 1更早的评论

Yokuna 2022-10-22

Thanks. Updated.

Torsten 2022-10-22

编辑：Torsten 2022-10-22

What is the function "lamda" for ? "fminsearch" searches for a value of "lambda" such that the objective function becomes minimal. There is no reason to vary or prescribe lambda on your own. fminsearch will do this for you.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Steven Lord 2022-10-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1832828-calling-function-in-fminsearch#answer_1081288

在 MATLAB Online 中打开

If you mean you want to call fminsearch repeatedly with different initial points, call it in a loop.

f = @sin;
initialPoints = (0:0.5:10).';
solutions = zeros(size(initialPoints));
for k = 1:numel(initialPoints)
    x0 = initialPoints(k);
    solutions(k) = fminsearch(f, x0);
end

For ease of display I'll put the points in a table.

results = table(initialPoints, solutions)
results = 21×2 table
initialPointssolutions______________________

           0          -1.5708 
         0.5          -1.5708 
           1          -1.5708 
         1.5          -1.5708 
           2           4.7124 
         2.5           4.7124 
           3           4.7124 
         3.5           4.7124 
           4           4.7124 
         4.5           4.7124 
           5           4.7124 
         5.5           4.7124 
           6           4.7124 
         6.5           4.7124 
           7           4.7124 
         7.5           4.7124 

2 个评论
显示无隐藏无

Yokuna 2022-10-22

Thanks! But I need the initial value as the value obtained in previous iteration of solution.

So initialPoints are not predefined but calculated in each iteration.

Steven Lord 2022-10-22

在 MATLAB Online 中打开

Okay. I'll use randi to compute my initial points, but you could use another function. In fact you could (except when whichTrial is 1) look at results(whichTrial-1) to use the results from the previous run as x0 for this run.

ntrials = 10;
initialPoints = zeros(ntrials, 1);
results = zeros(ntrials, 1);
f = @sin;
for whichTrial = 1:ntrials
    x0 = randi([-10 10]); % You could compute this however you want
    initialPoints(whichTrial) = x0; % Record it
    results(whichTrial) = fminsearch(f, x0); % Compute with it
end

Now show the results table.

t = table(initialPoints, results)
t = 10×2 table
    initialPoints    results
    _____________    _______

         -5           -7.854
         -8           -7.854
          2           4.7124
         -6           -7.854
         10           10.996
         10           10.996
          9           10.996
          1          -1.5708
         10           10.996
         -3          -1.5708