When I use Classify on a dataset with inputs that are matrices with different row sizes (i.e., sequences), and I check the internals of the ClassifierFunction by running c[[1]], I see that Mathematica automatically applies some preprocessor that is called NumericalSequenceExtractFeatures. I hovered over it to see what it is doing, but it is unclear to me what kind of feature extraction it is performing. I was wondering what exactly does it do? See the attached image for a screenshot of what I am talking about.


The input type that Mathematica assigns to my dataset is NumericalVectorSequence, and it seems that this preprocessing step is done whenever you have a dataset of vector sequences, where the sequence length is not constant.

enter image description here

  • 1
    $\begingroup$ "NumericalSequence" - that's the only reference I could find in the docs. It's a feature extraction mode when using variable length sequences in the training data but there's not much information on how it achieves this. $\endgroup$
    – flinty
    Mar 30 at 23:52
  • $\begingroup$ @flinty thanks for the comment. The input type that Mathematica assigns to my data is NumericalVectorSequence. It appears that Mathematica applies that preprocessing step to any dataset that consists of vector sequences of varying length. But it is still not clear what it is exactly that the preprocessing step does. $\endgroup$
    – sepehr78
    Mar 31 at 10:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.