how i can vectorize this code using bsxfun or meshgrid?

Hi,

I am trying to vectorize this code, but I dont know how to start it, I read the help files for both bsxfun and meshgrid but no luck. Can someone please guide how to look at this rather than giving the code. My code is working for me but i want to improve it.

s1= 50;
s2=100;
E_=zeros(s2,s2);
grid_=linspace(-2.5,2.5,s2);
tau=linspace(-pi,pi,s1);
x=cos(tau);
y=sin(tau);
cur=myfun(x,y); %returns vector
for k=1:s2
    for m=1:s2
        for n=1:s1
            R= abs(sqrt((grid_(m)-x(n))^2+(grid_(k)-y(n))^2));
            u_(n)= besselj(0,R)*cur(n);
        end
        E_(k,m)=abs(trapz(tau,u_));
    end
end

Thanks

Turker

Answer 1

[Jan]

You can replace the inner loop by:

R  = abs(sqrt((grid_(m)-x) .^ 2 + (grid_(k)-y) .^ 2));  % [EDITED]: ^ to .^
u_ = besselj(0, R) .* cur;

I guess, that the abs() is not required.

[ttopal]

Thanks Jan, x and y are array. And I a calculating the distance between points one by one. When we change it like this, does R become an array?

[Jan]

Yes, R has the same size as x and y, and u_ has also. At least if you run the improved version which uses the elementwise squaring .^2. See [EDITED]

[ttopal]

Hi Jan, it returns error as;

Assignment has more non-singleton rhs dimensions than 
non-singleton subscripts

Error in testfields (line 121)
        E_(k,m)=abs(trapz(tau,u_));

[ttopal]

R  = abs(sqrt((grid_(m)-x) .^ 2 + (grid_(k)-y) .^ 2))'; %transpoze added.
u_ = besselj(0, R) .* cur;

Now it works! Thank you Jan!