This repository is used to test the effect of the second stage model (arbitration) and window length (scope) on the EEG abnormal classification task.
batch_test_hyperparameters.default.py: A template for hyperparameters that need to be tested in batches
train_and_eval_config.default.py: Template for hyperparameters tested individually
train_and_eval.py: Train and test the first-stage model
final_decision.py: Train and test the second-stage model
util.py: Helper functions
vit.py, hybrid_1.py, tcn_1.py: model file
tueg_labels.csv: labels of TUEG
results_boxplot.py Plot results
1, Download the TUAB dataset https://isip.piconepress.com/projects/tuh_eeg/downloads/tuh_eeg_abnormal/
2, Create batch_test_hyperparameters.py and train_and_eval_config.py based on batch_test_hyperparameters.default.py and train_and_eval_config.default.py
3, Run the train_and_eval.py to generate result.csv and training_detail.csv
4, Run the final_decision.py to generate decision_result.csv