Unique Function based on 2 columns [Instead of rows]

Suppose, I have a cell array, a, with contents as such:

a{1}=[1 3 4 5;
      3 3 4 5;
      5 5 4 5
      2 4 2 6;
      6 5 2 6
      7 2 3 1;]

How can I apply the 'unique' function on 2 columns [column 3 and 4] such that they will return the value '3' and '2'. [Since there are 3 duplicates for the pair 4,5 and 2 duplicates for the pair 2,6.]

Solution proposed:

diff(find([true; any(diff(a{1}(:,3:4)),2); true]))

-----------------------------------------------------------------------

However, when the arrangement of the contents is changed, for eg.

a{1}=[  1 3 4 5;
        3 3 2 6;
        5 5 4 5
        2 4 3 1;
        6 5 2 6
        7 2 3 1;]

The code doesn't seem to identify the similarity/duplicated for column 3 and 4 [Corresponding rows: 1 & 3; 2 & 5; 4 & 6]

[Cedric]

No, this code assumes blocks of similar rows (for cols 3 and 4), as it seems to be the case in your first example.

Answer 1

[Azzi Abdelmalek]

b=a{1}
c=b(:,3:4)
[~,idx]=unique(c,'rows','first','stable')
out=b(idx,:)

[RDG]

Thank you Azzi for the concise steps. However, it doesn't give the count of the number of duplicates. [In this case, 2,2,2 for each pair of duplicates]. Is there a way to count the number of duplicates?

[Cedric]

There is certainly a simpler way, but if you add a 3rd output parameter to the call to UNIQUE

 [~,idx,cnt] = unique(c, ... )

you can get effective counts with the following

 >>  cnt = accumarray(cnt, ones(size(cnt)))
 cnt =
     2
     2
     2

[RDG]

This problem which I'm facing has exposed me to so many new commands in Matlab and logic which I never thought of before. Thank you for your great insight!

[Cedric]

Yep, in some sense it is easier to understand how to solve an ODE than to understand how to perform data manipulation efficiently ;-)