Regarding improving efficiency of the code using fzero and int (integral) function in for loop

Hi ,

I am working on data generation based on user input values using probability distribution functions. I have to repetatively solve an equation with single variable , which sometimes involves integration too. I used the 'fzero' and' int' function in the code in the 'for' loop. I am facing 2 problems here:

1. The fzero function gets stuck in a loop , as it cannot find a real and finite solution
2. In case , even if it finds a solution , the program takes long time to generate the data.

I have attached the code for your reference. Run the 'f_example.m' file to check the code.

Answer 1

[darova]

I tried to plot your function

c1 = .9;
c2 = 0.7;
c3 = 0.6;

x = -1:0.1:1;
f = @(x) c1 - exp(c2*x).*(1-c3*x);
plot(x,f(x))

As you can see you can have only 2 roots max

So this piece of a code needs to be improved

while (isnan(k(i1)) || isinf(k(i1)) )
    x_sol_0=rand;
     if isreal(fun(x_sol_0))
        k(i1) = fzero(fun,x_sol_0);
     end
end

Especially this line

x_sol_0=rand;

Maybe this way

x_sol_0=-x_sol;

Are you sure there are always roots that meet condition?

[Yasho Bharat Boggarapu]

x_sol_0=rand;

I use this to change the initial value for fzero to solve , if it encounters any NaN or inf value in the previous line. It should have the roots always. I will try your way and see , if it helps.

[Yasho Bharat Boggarapu]

while (isnan(k(i1)) || isinf(k(i1)) )
    x_sol_0=rand;
     if isreal(fun(x_sol_0))
        k(i1) = fzero(fun,x_sol_0);
     end
end

This one is used because the line before( which I have show below )gives NAN or inf values sometimes. So I have checked that condition and tried to change the initial point for fzero to solve.

        fun = @(kt)(kt_u*lamda*v*(1+erf(x(i1)/sqrt(2)))/(2*c))+1 - (exp(kt*lamda) *(1-v*kt));
        k(i1,1) = fzero(fun,x_sol_0);
        

But some how , it gets stuck in the above while loop sometimes.

[Yasho Bharat Boggarapu]

Hi darova,

x_sol_0 =-x_sol_0

This still doesnt work. It didnot help.

[darova]

What about fsolve?