Find non-unique elements of an array in Matlab

Question

If I have an array [1 2 3 4 3 5 6 7 8 7], I'd like to find the list of non-unique entries: [3 7]. I fail to find a simple way to do it. Any idea?

Update: I'd like to have a universal solution, which would also work with cell array of strings.

possible duplicate of http://stackoverflow.com/questions/5385651/determining-the-number-of-occurrences-of-each-unique-element-in-a-vector — gregswiss, Sep 23 '15 at 11:12
@gregswiss: I need a list of duplicated elements rather than how often they are encountered. Besides, the solutions in the linked question are not applicable to non-numeric arrays. — texnic, Sep 23 '15 at 12:25

score 5 · Accepted Answer · answered Sep 23 '15 at 10:43

5

If A has length n, you can find the indices in A of the first occurrence of each entry and remove them from A:

A = [1 2 3 4 3 5 6 7 8 7];
n=length(A);
[~,IA,~] = unique(A);
out = unique(A(setdiff((1:n),IA)))

answered Sep 23 '15 at 10:43

Steve

1,579
10
23

Divakar · Answer 2 · 2015-09-23T11:06:50.137

One approach with unique and histc -

[unqA,~,id] = unique(A);
out = unqA(histc(id,1:max(id))>1)

Or use accumarray in place of histc -

out = unqA(accumarray(id(:),1)>1)

Or use bsxfun -

out = unqA(sum(bsxfun(@eq,id(:),1:max(id(:)).'))>1)

Sample runs -

1) Numeric arrays case -

>> A
A =
     6     3     7     7     4     3     8     5     2     3     1
>> [unqA,~,id] = unique(A);
>> unqA(histc(id,1:max(id))>1)
ans =
     3     7
>> unqA(accumarray(id(:),1)>1)
ans =
     3     7
>> unqA(sum(bsxfun(@eq,id(:),1:max(id(:)).'))>1)
ans =
     3     7

2) Cell arrays case -

>> A = {'apple','banana','apple','mango','ball','cat','banana','apple'};

>> [unqA,~,id] = unique(A);
>> unqA(histc(id,1:max(id))>1)
ans = 
    'apple'    'banana'
>> unqA(accumarray(id(:),1)>1)
ans = 
    'apple'    'banana'
>> unqA(sum(bsxfun(@eq,id(:),1:max(id(:)).'))>1)
ans = 
    'apple'    'banana'

Nowadays, `histcounts` is recommended instead of `histc`. – Rantanplan Oct 17 '22 at 16:52 — Rantanplan, Oct 17 '22 at 16:52

score 2 · Answer 3 · answered Jul 20 '16 at 17:25

2

x=[1 2 3 4 3 5 6 7 8 7];
y=x;
[~,ind,~]=unique(y);
y(ind)=[];

y is the non-unique entries.

answered Jul 20 '16 at 17:25

Mirza

31
4

Although this code may be help to solve the problem, providing additional context regarding _why_ and/or _how_ it answers the question would significantly improve its long-term value. Please [edit] your answer to add some explanation. – oɔɯǝɹ Jul 20 '16 at 19:16

score 0 · Answer 4 · answered Sep 23 '15 at 11:08

Since you ask for a more generic solution, here is one that should be easily adaptable to other data types. Compared to others it is also a O(n) solution - but the drawback is the slow Matlab looping on large array of elements...

A = [1 2 3 4 3 5 6 7 8 7];
dupes = [];

map = containers.Map('KeyType', class(A), 'ValueType' , 'logical');
for i=1:numel(A)
    if map.isKey(A(i))
        dupes = [dupes A(i)];
    else
        map(i) = true;
    end        
end

score -1 · Answer 5 · edited Sep 23 '15 at 10:57

-1

Another approach using sort and diff:

As = sort(A);
out = unique(As([false diff(As)==0]));

edited Sep 23 '15 at 10:57

Steve

1,579
10
23

answered Sep 23 '15 at 10:19

Bentoy13

4,886
1
20
33

3

It seems more than two occurrences of a number would give duplicated entries in the output, unless I missed something. – Divakar Sep 23 '15 at 10:36
@Divakar No, you don't miss anything, you're absolutely right. Thanks Steve to have edited my answer. – Bentoy13 Sep 23 '15 at 10:58

Find non-unique elements of an array in Matlab

5 Answers5