Simon Kern

The biggest problem in my opinion is, that there is no Gold standard and that current sleep staging is just a sub-optimal practice kept alive since the 60s. Probability scores...

No, see here for details: https://physionet.org/pn4/sleep-edfx/ If you find a database with 4 scorers per record, let me know.

100% accuracy is certainly overfitting. With sleep data, if you get past 90% accuracy on your training set, you can be sure that you were overfitting. Also numbers on the...

If this is on hold out data I assume there is some error in your setup. Reaching 100%, especially with sleep data, is close to impossible, due to the limit...

Thanks! Good one.

@mayeaux theoretically no problem, technically currently not possible. The data here is trained on different systems using different electrode positions. In theory it would be no problem to retrain data...

Ah, the good old Zeo. Although it is a great device I would not base any assumptions on its output on non-sleep recordings. Algorithms are often very fragile and even...

Just to add up on that: I don't think there are high similarities between deep sleep and Yoga. As said before, it's just an algorithm and it will misclassify your...

Thanks! The (preliminary) pre-trained weights that I supplied need all three channels to work correctly. For usage of only one channel (EEG), you would need to re-train a network. It...

It is possible to just remove the channels not used. However, it is necessary to re-train the network. Although training one network to be able to use a variable subset...