CNN-LSTM-and-multihead-attention-for-classifying-electroencephalography-EEG-data

KO | EN

gitlite — search

#python #java #python3 #arduino #golang #machine-learning #rust #html #flask #javascript #seismology #nodejs

CNN-LSTM-and-multihead-attention-for-classifying-electroencephalography-EEG-data

★ 3 Open GitHub ↗

No description available.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# CNN-LSTM-and-multihead-attention-for-classifying-electroencephalography-EEG-data

We implement the vanilla CNN, CNN+MHSA, and CNN+LSTM architectures for the EEG decoding task, and train these models on the data of subject 1 and all subjects to evaluate their relative performances. The testing accuracies of the vanilla CNN and CNN+MHSA achieve the desired 70% threshold, but CNN+LSTM does not. In addition, we analyze the models with respect to classification accuracy as a function of time. Given a training time limit of 100 epochs, we find that CNN+MHSA requires 77 epochs, the longest among the three models, to reach testing accuracy of 70%, with its performance possibly continuing to increase after epoch 100. In contrast, the vanilla CNN converges to the testing accuracy benchmark of 70% early on epoch 36, whereas CNN+LSTM fails to even reach the threshold by the epoch 100 cutoff. Additionally CNN and CNN+LSTM models both show early significant divergences between training accuracy and validation accuracy.

CNN-LSTM-and-multihead-attention-for-classifying-electroencephalography-EEG-data

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README