Aside from Convolution Neural Networks, are there any other methods that allow for classification of Tensors? My observations consist of multi-dimensional tensors with height of 1, where each channel corresponds to a particular time-series and am wondering how I can effectively classify the tensors, taking into account the relationship between the time-series.

  • $\begingroup$ You can flatten the tensor and run the usual machine learning methods on the vector: random forest, kNN, SVM, logistic regression, etc. $\endgroup$ – Dave Apr 20 at 1:27

Recurrent neural networks might be an option. These have the capability of keeping information in a time series and are used for signals in time domain, for applications such as speech recognition etc... LSTM is a variant of these which is also widely used.

| cite | improve this answer | |
  • 1
    $\begingroup$ Very good point, but my issue is that I don't have one-to-one labelling, but rather one label for the entire tensor. I was under the impression that RNNs require labels at every time-step. Is there any other method that comes to your mind? $\endgroup$ – mamafoku Aug 18 '17 at 15:43
  • $\begingroup$ Can you use the same label at each time step? If the label is cat then it’s always a cat, right? $\endgroup$ – Dave Apr 20 at 1:22

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.