Look for minimum value in array

Hello,

I have a function like this:

i=1;
Min=ones(1,200);

while i>=200

min_y = min(y(:,i));
pos_y = find(y(:,i)==min_y);
phi0s = x(pos_y ,i);

Min(1,i) = phi0s;

i=i+1;
end

Its job is to find a min value of y and as a results give a corresponding x value.

x and y are arrays with a dimension of n x 200 - let say that I stored results of 200 experiment in those two arrays and now I need to find smallest y value in each column and corresponding x.

Can I improve it somehow or even better run it without a loop? I will be grateful for any tips!

Answer 1

[Rik]

Loops aren't slow. You generally only need to remove them if there are native functions that accept array inputs, which happens to be the case here:

%generate random data
n=4;
x=rand(n,200);
y=rand(n,200);

%your corrected code
i=1;
Min=ones(1,200);
while i<=200%why aren't you using a for-loop here?
    min_y = min(y(:,i));
    pos_y = find(y(:,i)==min_y);
    phi0s = x(pos_y ,i);
    Min(1,i) = phi0s;
    i=i+1;
end

%vectorized
[~,idx]=min(y,[],1);
ind=sub2ind(size(x),idx,1:size(y,2));
Min2=x(ind);

isequal(Min2,Min)

[Mateusz Brzezinski]

Thanks a lot!

Yes, for loop will also fit (even better).

[Mateusz Brzezinski]

Ok but what if I have two y values that are found as min - let say that I have in one column 4, 3, 1, 1?

How to overcome it?

[Image Analyst]

And the 1 can be omitted

Min(i) = phi0s;

And you should probably use a for loop like he suggested, and use "col" or "column" for the column index rather than i which is the imaginary variable sqrt(-1). Or use the vectorized approach. It's more compact and cryptic than a for loop though. So use whatever you're comfortable with. For loops will be fast, particularly if there are only a microscopic 200 iterations.

[Mateusz Brzezinski]

Ok, thanks I will do that

Answer 2

[David Hill]

[Y,idx]=min(y);
X=x(length(y)*(0:length(idx)-1)+idx);