Is there a one-liner to find the minimum of unique entities from a matrix of multiple instances of each item?

Question

Say, I have a matrix A with 2 columns - column 1 contains item ID and column 2 contains its weight:

A = [
3   5
2   3
2   5
1   4
3   4
2   6
1   9
3   2 ];

I want the output as follows:

items = [
1   4
2   3
3   2];

The code I would write to do this would be:

items(:,1)=unique(A(:,1));
for i=1:size(items,1)
    temp=A(A(:,1)==items(i,1),:);
    items(i,2)=min(temp(:,2));
end

The items matrix is the required output here.

I was wondering if there is a one-liner code that does this in MATLAB.

Answer 1

assuming your item IDs are integers starting from one, you can use accumarray:

accumarray(a(:,1), a(:,2), [], @min);

Answer 2

Use sortrows:

B = sortrows(A);
A_min = B([true; diff(B(:,1))~=0], :)

A_min =

     1     4
     2     3
     3     2

sortrows sorts the matrix in ascending order based on the first column, then the second. Use this new sorted vector and extract the first row, along with all rows where the first column changes (1, 2, 3).

[true; diff(B(:,1))~=0] is a vector containing boolean [true false false true ...]. true indicates where the elements in the first column change. This is used as a logical map to index the original array.

You should know that:

x = [1 2 3 4];
idx = [true false false true];
x(idx) = 
     1    4

Answer 3

If you have an access to image processing toolbox you can do this:

[unique(A(:,1)),[regionprops(A(:,1),A(:,2),'MinIntensity').MinIntensity]']