Identify the mode in each column and eliminate them

I am curious to see if the following could be done without loops.

I have a 7-by-N matrix with all integers. In each column of the matrix, there is one integer that appears 4 time (hence the most frequent integer).

This there any way to eliminate the most frequent integers in every column and to output the rest of the integers in a 3-by-N matrix without using loops?

Example: a = [1 1 2 2 2 2 3; 1 2 2 2 2 3 3; 2 2 2 2 3 3 3]';

Output should be [1 1 3; 1 3 3; 3 3 3]'

The order of numbers in each column could be randomized unlike in my example. Thank you very much and no this is not a homework problem.

[Han]

Edit: I used number 1,2 and 3 just for my example. The integers could be different. There could be different "mode" numbers (the most frequent numbers) in different columns.

Answer 1

[Andrei Bobrov]

reshape(a(ismember(a,[1,3])),[],size(a,2));

or

hankel([1 1 3],[3 3 3]);

ADD

s = size(a);
b = ones(s);
idxl = fliplr(triu(b,1) + tril(b,diff(s))) > 0;
out = reshape(a(idxl),[],s(2));

[Han]

Number 1,2 and 3 only appeared in my example. It does not mean they should be the ones you use in your code.

[Andrei Bobrov]

See ADD part in my answer.

Answer 2

[Iain]

 a = a';
 out = reshape(a(a~=2),[],3)';

That will work for your example, but not necessarily what you're dealing with.

[Han]

Number 1,2 and 3 only appeared in my example. It does not mean they should be the ones you use in your code.

Answer 3

[Roger Stafford]

Let 'a' be an array in which one element of each row is repeated the same number of times for all rows with no other element in each row occurring as often as this.

 b = a.';
 b = reshape(b(bsxfun(@ne,mode(b),b)),[],size(b,2)).';