Why sequentialfs changes the number of selected features every time when you run the code...?

Hi, I am using the sequentialfs function for the feature selection. But when i run the code multiple times the number of features changes. I would like to know why this happens and how to avoid it.

[Ritz]

It might be due to the radomly chosen training and test sets by cvpartition, if you are using kfold cross validation for the criterion function, which is generally preferred while using sequentialfs. The feature-set should not change very much across different runs, few may be different, but more or less it should be stable. In that case, you can report the most frequent feature across multiple runs.

Hope it helps !!