How to apply looping in iteration process ?

Hi everyone,

I required to apply diffrent initail guess for iterative solutions. For a single solution, the following script works perfectly.

L = readmatrix('R_bounds_0.01.csv'); (coloumn matrix with 1986 rows)

for i=1:length(L);
f = @(n)((n+1)*log(n+1)-n)/n^2 - L(i);
n0 = 100; % Initial guess
n(i) = fzero(f,n0 );
disp(n );
end
nn=n';
csvwrite('N_lower_100.csv',nn);

However, i need to try with different values of initial guesses [0, 10, 100, 1000] etc. Someone suggets me a modified script but that is still not working,

A = readmatrix('R_mean_value_0.01.csv');
m = length(A) ; 
IG = [0, 10, 100, 1000] ;    % initial guess 
n = length(IG) ;
nn = zeros(m,n) ; 
for i = 1:n
    n0 = IG(i) ; 
    for j=1:m
        f = @(n)((n+1)*log(n+1)-n)/n^2 - A(j);    
        nn(i,j) = fzero(f,n0 );
        disp(nn(i,j) );
    end
end
csvwrite('N_mean_10000.csv',nn);

May someone suggest me any possibel solution.

Thank you!

[Mathieu NOE]

hello

can you share the csv file as well (R_mean_value_0.01.csv)

[Andi]

Sure, Attached with main question ..

[Torsten]

Someone suggets me a modified script but that is still not working,

Do you mean that you get a syntax error message or that fzero has problems finding a solution ?

[Andi]

There is no syntax error, but fzero problem.

[Torsten]

The implementation is correct.

I think the task of the assignment was to show you that the rootfinder may fail if the initial guess is far off. So there is nothing wrong with your code if fzero does not converge in all test cases.

[Andi]

I agreed, sometime the initial guess is a bit far from the solution that why i used different initial guess. But that not works.

Answer 1

[Walter Roberson]

If your initial guess is not between about -1 and about +40 then you might not get a solution.

filename = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/930814/R_mean_value_0.01%20.csv';

A = readmatrix(filename);

syms N
f = @(n)((n+1)*log(n+1)-n)/n^2;
baseeqn = f(N);
fplot(baseeqn, [-1.5 40])
hold on
minA = min(A)

minA = 0.0706

maxA = max(A)

maxA = 0.9681

yline([minA, maxA]);
ylim([minA-1/4, maxA+1/4])
hold off

sol1 = vpasolve(baseeqn - A(1), 0.5)

sol1 = 

15.859241089220361924912561179785

fplot(baseeqn, [32 40]); hold on; 
yline([minA]);
hold off

[Andi]

Despite of chaniging the initial guess, still got same error

A = readmatrix('R_mean_value_0.01.csv');
m = length(A) ; 
IG = [-1.5, 1, 10, 40] ;    % initial guess 
n = length(IG) ;
nn = zeros(m,n) ; 
for i = 1:n
    n0 = IG(i) ; 
    for j=1:m
        f = @(n)((n+1)*log(n+1)-n)/n^2 - A(j);    
        nn(i,j) = fzero(f,n0 );
        disp(nn(i,j) );
    end
end

% Error 
Error using fzero
Initial function value must be finite and real.

Error in untitled (line 10)
        nn(i,j) = fzero(f,n0 );
        
        

I think initial guss cant be negative, therefore i used [0, 1, 10, 100] as an initial guess.

[Walter Roberson]

You use log(n+1) . n+1 would be negative for n < -1 and that would give you a complex log. So your n values can never be <= -1 .

You can use a vector of two values for the second parameter to fzero() to prevent fzero from guessing that a number < -1 might perhaps be acceptable.

[Andi]

The following script works but i unable to process teh output i dont knwon why (e.g. unable to write in a csv file)

A = readmatrix('R_mean_value_0.01.csv');
m = length(A) ; 
IG = [0.0001,0.001, 0.01, 0.1, 1, 10, 100, 1000] ;    % initial guess 
n = length(IG) ;
nn = zeros(8,n) ; 
for i = 1:n
    n0 = IG(i) ; 
    for j=1:m
        f = @(n)((n+1)*log(n+1)-n)/n^2 - A(j);    
        nn(i,j) = fzero(f,n0 );
        disp(nn(i,j) );
    end
end

%SS=nn';

csvwrite('NNNN.csv',nn);

%TTT = (SS(:,1), SS(:,2), SS(:,3), SS(:,4), SS(:,5), SS(:,6), SS(:,7), SS(:,8));

[Walter Roberson]

As an outsider.... I do not know that your current directory is one that you have write access to.

[Andi]

If i run the same scipt using one initail guess, i can write the output in csv but when use for loop then i cant. This seems that tehre is no issue of directory.

[Mathieu NOE]

hello

a very minor bug in the first code posted is the initialization of nn is done with the wrong dimensions (flip m and n)

I have no problel running the corected code - attached the output file

of course the IG must contains only values strictly above -1 (otherwise log(n+1) is complex) and different from zero (division by zero)

A = readmatrix('R_mean_value_0.01.csv');
m = length(A) ; 
IG = [0.0001,0.001, 0.01, 0.1, 1, 10, 100, 1000] ;      % initial guess (n> - 1) and (n # 0)
n = length(IG) ;
nn = zeros(n,m) ; % mod here
for i = 1:n
    n0 = IG(i) ; 
    for j=1:m
        f = @(n)((n+1)*log(n+1)-n)/n^2 - A(j);    
        nn(i,j) = fzero(f,n0 );
        disp(nn(i,j) );
    end
end
csvwrite('N_mean_10000.csv',nn);