Question

If I had a matrix A such as:

63    55    85    21    71
80    65    85    48    53
55    60    93    71    66
21    65    40    33    21
61    90    80    48    50

… and so on how would I find the minimum values of each column and remove those numbers from the matrix completely, meaning essentially I would have one less row overall.

I though about using:

[C,I] = min(A);

A(I) = [];

but that wouldn’t remove the necessary numbers, and also reshape would not work either. I would like for this to work with an arbitrary number of rows and columns.

Answer 1

A  = [
63    55    85    21    71
80    65    85    48    53
55    60    93    71    66
21    65    40    33    21
61    90    80    48    50
];

B = zeros( size(A,1)-1, size(A,2));
for i=1:size(A,2)
    x = A(:,i);
    maxIndex = find(x==min(x(:)),1,'first');
    x(maxIndex) = [];
    B(:,i) = x;
end
disp(B);

Another vectorized solution:

M = mat2cell(A,5,ones(1,size(A,2)));
z = cellfun(@RemoveMin,M);
B = cell2mat(z);
disp(B);

function x = RemoveMin(x)
    minIndex = find(x==min(x(:)),1,'first');
    x(minIndex) = [];
    x = {x};
end

Another solution:

[~,I] = min(A);
indexes = sub2ind(size(A),I,1:size(A,2));
B = A;
B(indexes) = [];
out = reshape(B,size(A)-[1 0]);

disp(out);

Personally I prefer the first because:

1. For loops aren’t evil – many times they are actually faster (By using JIT optimizer)
2. The algorithm is clearer to the developer who reads your code.

But of course, its up to you.