Newbie to Matlab: How to find duplicates within a cell of arrays, arrays have varying sizes

Hi,

I have a 100 x 1 cell containing arrays of varying length from 5 to 8. Here's an example:

A = [1x7 double] [1x6 double] [1x6 double] [1x7 double] [1x5 double]

Each of these arrays looks something like [1 5 20 12 25 27 1]. Is there a fast way to identify any duplicates within the entire cell and delete them so you're left with only unique arrays? I've been searching on this site but can't seem to find a way to deal with varying length arrays.

Thank you so much for your help!

Newbie to Matlab

Answer 1

[Teja Muppirala]

Here is a one rather exotic way of doing it:

% Step 1: Convert to strings
Aunique = cellfun(@(x)char(typecast(x,'uint8')), A, 'unif',0);

% Step 2: Call the UNIQUE command
Aunique = unique(Aunique);

% Step 3: Reconvert to doubles
Aunique = cellfun(@(x) typecast(uint8(x),'double'),Aunique,'unif',0)

[andrew]

Wow, very exotic but it works! thank you

[andrew]

Okay, as an extension to this problem, if I have another 100 x 1 matrix that is related to my first cell (ex. this second cell would be the values associated with the first cell), then is there a good way using what you've done to ensure I delete the row in the second cell at the same time I delete the row in the first?

For example, if I have:

A = [1x7 double] [1x6 double] [1x6 double] [1x7 double] [1x5 double]

and values = [75; 55; 50; 45; 75]

When I delete the [1x7] in A assuming these are the same, how can I make sure I delete the corresponding 75 in the values array

Answer 2

[Walter Roberson]

A( any(triu(bsxfun(@isequal, A, A.'),1)) ) = [];

[andrew]

Thank you so much, but I keep getting the error:

'Operands must be numeric arrays'

when I try to use your code on the following cell:

A = {[1 5 20 12 25 27 1];[1 12 20 25 27 1];[1 20 25 27 1];[1 5 20 12 25 27 1]}

Am I missing something here? Thank you

Andrew

[Walter Roberson]

There might be a version difference: the current documentation does indicate that cells are allowed.

You could use

A( any(triu(0+bsxfun(@(J,K) isequal(A{J},A{K}), (1:length(A)).', 1:length(A)),1)) ) = [];

[andrew]

Thanks again, now I'm getting the error 'invalid output dimensions' maybe I'm using an old version?