How to create indices for the k-fold cross-validation?
조회 수: 9 (최근 30일)
이전 댓글 표시
Hi All,
I'd like to create indices for the k-fold cross-validation using
indices = crossvalind('Kfold',Labels,k);
The "Labels" is a 1-by-1000 cell array which contains 1000 cells, as follows
Labels(1: 10) = 1×10 cell array
Columns 1 through 9
{1×1 cell} {1×1 cell} {1×1 cell} {1×1 cell} {1×1 cell} {1×1 cell} {1×1 cell} {1×1 cell} {1×1 cell}
Column 10
{1×1 cell}
Labels{1}= 1×1 cell array
{'2'}
I got this error
Undefined function 'crossvalind' for input arguments of type 'cell'.
So, I took the following steps to solve the issue
Label_1 = cellfun(@(x) str2double(x), Labels, 'UniformOutput',false);
Label_double = [Label_1{:}]
indices = crossvalind('Kfold',Label_double,10);
And now I am dealing with the following error
Undefined function 'crossvalind' for input arguments of type 'char'.
I'm not sure why I get this error. any idea? Thanks!
댓글 수: 3
Megumi Fukuda
2021년 6월 9일
Could you upload your label variable (you can upload .mat file) so that I can have a look? I assume your label is what-I-call 'nested' and that's why you cannot use crossvalind.
채택된 답변
Megumi Fukuda
2021년 6월 10일
indices = crossvalind('Kfold',Labels,k);
I am sure this does not work, as the cells in the variable Label are nested and cannot be used for crossvalind without converting the variables.
I have tested this code with MATLAB 2021a and this code looks fine. Would you like to try your code again?
% clear the workspace
load('Label.mat'); % the mat file you uploaded
Label_1 = cellfun(@(x) str2double(x), Giant_bpm, 'UniformOutput',false);
Label_double = [Label_1{:}];
indices = crossvalind('Kfold',Label_double,10);
(... and here is another implementation I came up with. This returns the label as char type)
load('Label.mat');
Giant_bpm_updated = cellfun(@cell2mat, Giant_bpm(:), 'UniformOutput', false);
indices = crossvalind('Kfold',Giant_bpm_updated,10);
Hope this helps!
추가 답변 (0개)
참고 항목
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!