cvpartition with specified indices

I want to manually specify the indices in cvpartition for leave-one-out cross validation with 5 subjects. Right now the data is split into 5 equal parts if I set "KFold" to 5.

train = [1:50];

cvpt = cvpartition(train,"KFold",5)

K-fold cross validation partition

NumObservations: 50

NumTestSets: 5

TrainSize: 40 40 40 40 40

TestSize: 10 10 10 10 10

Now I want to manually specify which indices are used for training and which for testing. Lets say the indices of my 5 train and test sets are:

Test1_idx = [1:8]; Train1_idx = [9:50];

Test2_idx = [9:21]; Train2_idx = [1:8 22:50];

Test3_idx = [22:31]; Train3_idx = [1:21 32:50];

Test4_idx = [32:44]; Train4_idx = [1:31 45:50];

Test5_idx = [45:50]; Train5_idx = [1:44];

How can I specify this in Matlab before I fit machine learning model?

The cvpt should then look like this:

NumObservations: 50

NumTestSets: 5

TrainSize: 42 37 40 37 44

TestSize: 8 13 10 13 6

Answer 1

[Drew]

As of R2023b, users can manually specify the indices in a cvpartition with the syntax

c = cvpartition("CustomPartition",testSets) 

The details, including the possible formats for the "testSets" specification, are in the R2023b+ cvpartition doc https://www.mathworks.com/help/stats/cvpartition.html. This new feature is also mentioned in the R2023b SMLT release notes at https://www.mathworks.com/help/stats/release-notes.html.

If this answer helps you, please remember to accept the answer.