1

I was trying to compare the accuracy results of CNN model and combined CNN-SVM model for classification. However I found that CNN model have better accuracy than combined CNN-SVM model. Is That correct or it can happen?

root
  • 59
  • 1
  • 7

1 Answers1

1

It depends on a large number of factors , but yes if the underlying data is image - cnn have proven to deliver better results.

Prachi
  • 402
  • 1
  • 5
  • 10
  • Yes, actually the input data was EEG signals that have been transformed to spectrograms that will be fed as input to CNN. For some cases the CNN-SVM gives better results however in other cases it was clearly bad than CNN model. So I was wondering about the reasons of this?? – root Nov 14 '19 at 00:36
  • In most cases, CNN will be better for such data sets and if you dig deep, the data might fit Resnet and other cnn variants even better. Later may be an ensemble of svm and other variants can be researched to optimise results. Thanks. – Prachi Nov 14 '19 at 04:39