I am trying to use a network for classification. This network works very well on the author's example data, but doesn't work on new data.

Currently, I am using the popular EEG Motor Movement/Imagery Dataset, where other people arrived at ~80% accuracy. I got this accuracy with brain decode (CNN).

Now, I am trying to classify this data using EEGLearn (RNN-CNN), but I'm still on the chance level. Do you have any suggestions?

  • 1
    $\begingroup$ Hi and welcome to this community! Can you clarify this: "but I'm still on the chance level"? Do you mean that you're getting a very low accuracy? $\endgroup$ – nbro Oct 10 '19 at 2:38

The implementation of any neural network needs precise feature engineering on your dataset. Neural network is very tailored solution. In case of classification, braindecode uses SignalAndTarget class. This is where signal is mapped on x axis and target label is on y axis.

You have to check whether you're following the same mapping. EEG learn is using time series data and implementation is done in tensorflow.

My advise would be to relatively scale the data into time scale used in EEG method.

| improve this answer | |

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.