How to find the mode of a string array (no longer works in 2018b)

15 views (last 30 days)
I have discovered that the following code worked in Matlab 2018a but now throws an error in 2018b:
mode(["A","A","B"],2)
It would appear that the new version of matlab only allows computing the mode on numeric arrays. Is this a bug, or a known change to how this function works (I know that in this version, some of these functions now operate on multiple dimensions now)? And is there a quick workaround? So far I've created the following hack to operate on strings, though I expect it's pretty inefficient:
Y = strings(1,size(X,2));
for n = 1:size(Y,2)
C = categorical(X(:,n));
CC = categories(C);
[~,ind] = max(countcats(C));
if ~isempty(ind)
Y(n)=string(CC(ind));
end
end

Accepted Answer

Adam Danz
Adam Danz on 16 Nov 2018
Edited: Adam Danz on 16 Nov 2018
The documentation for both 2018a and 2018b indicate that the first input to mode() "can be a numeric array, categorical array, datetime array, or duration array." One workaround would be to change your string to a categorical vector.
This works in 2018b
mode(categorical( ["A","A","B"]))
ans =
categorical
A
If needed, you could convert back to a string:
string(mode(categorical( ["A","A","B"])))
  2 Comments
Anil Kamath
Anil Kamath on 19 Nov 2018
Thanks!
It's a shame this doesn't work using the native mode function on strings, as it means you can't quickly compute the mode in a 2D array along a single dimension without converting each slice into a categorical array one by one, and computing the mode of each categorical slice. It feels like a very natural interpretation for the mode of a string to be the string which occurs most frequently...

Sign in to comment.

More Answers (1)

Philip Borghesani
Philip Borghesani on 16 Nov 2018
Use a character vector instead of a string array:
mode(['A','A','B'],2)
% or
mode('AABBCCC',2)
I belive it was a bug that it worked with strings in R2018a. It has always worked with character vectors and sombody was a bit too helpful adding support for strings which can’t simply be treated as a numeric value.
  1 Comment
Anil Kamath
Anil Kamath on 19 Nov 2018
I'm afraid this won't work in my use-case as it has to work on multi-character strings rather than just scalar chars (I put a simplified example in my initial question). Adam's answer above looks like it works.

Sign in to comment.

Categories

Find more on Data Type Identification in Help Center and File Exchange

Products


Release

R2018b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!