This repository has been archived by the owner on Dec 7, 2021. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 23
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
25 changed files
with
926 additions
and
62 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -10,7 +10,7 @@ tuh_eeg_artifact: | |
|
|
||
| tuh_eeg_seizure: | ||
| echo "Request your access password at: https://www.isip.piconepress.com/projects/tuh_eeg/html/request_access.php" | ||
| rsync -auxvL [email protected]:~/data/tuh_eeg_seizure/v1.5.1 data/tuh_eeg_seizure | ||
| rsync -auxvL [email protected]:~/data/tuh_eeg_seizure/v1.5.2 data/tuh_eeg_seizure | ||
|
|
||
| eegmmidb: | ||
| wget -r -N -c -np https://physionet.org/files/eegmmidb/1.0.0/ -P data | ||
|
|
||
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1 @@ | ||
| logs* |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,230 @@ | ||
| #!/usr/bin/env python | ||
|
|
||
| # Ignore MNE and TensorFlow warnings | ||
| import warnings | ||
| warnings.simplefilter(action='ignore') | ||
|
|
||
| # Import TensorFlow with GPU memory settings | ||
| import tensorflow as tf | ||
| gpus = tf.config.experimental.list_physical_devices('GPU') | ||
| try: | ||
| for gpu in gpus: | ||
| tf.config.experimental.set_memory_growth(gpu, True) | ||
| except RuntimeError as e: | ||
| print(e) | ||
|
|
||
| # Import TensorBoard params and metrics | ||
| from tensorboard.plugins.hparams import api as hp | ||
| from tensorflow.keras.metrics import CategoricalAccuracy, Precision, Recall | ||
|
|
||
| # Import Spektral for GraphAttention | ||
| import spektral as sp | ||
|
|
||
| # Others imports | ||
| import os | ||
| import pickle | ||
| import numpy as np | ||
| from random import shuffle | ||
| from itertools import product | ||
| from networkx import to_numpy_matrix | ||
| from sklearn.model_selection import train_test_split | ||
| from tensorflow.python.keras.utils.np_utils import to_categorical | ||
|
|
||
| # Relative import PyEEGLab | ||
| import sys | ||
| from os.path import abspath, dirname, join | ||
|
|
||
| sys.path.insert(0, abspath(join(dirname(__file__), '../..'))) | ||
| from pyeeglab import * | ||
|
|
||
| def build_data(dataset): | ||
| dataset.set_cache_manager(PickleCache('../../export')) | ||
|
|
||
| preprocessing = Pipeline([ | ||
| CommonChannelSet(), | ||
| LowestFrequency(), | ||
| ToDataframe(), | ||
| MinMaxCentralizedNormalization(), | ||
| DynamicWindow(8), | ||
| ForkedPreprocessor( | ||
| inputs=[ | ||
| SpearmanCorrelation(), | ||
| Mean(), | ||
| Variance(), | ||
| Skewness(), | ||
| Kurtosis(), | ||
| ZeroCrossing(), | ||
| AbsoluteArea(), | ||
| PeakToPeak(), | ||
| Bandpower(['Delta', 'Theta', 'Alpha', 'Beta']) | ||
| ], | ||
| output=ToMergedDataframes() | ||
| ), | ||
| ToNumpy() | ||
| ]) | ||
|
|
||
| return dataset.set_pipeline(preprocessing).load() | ||
|
|
||
| def adapt_data(data, test_size=0.1, shuffle=True): | ||
| if isinstance(data, str): | ||
| with open(data, 'rb') as f: | ||
| data = pickle.load(f) | ||
| samples, labels = data['data'], data['labels'] | ||
| x_train, x_test, y_train, y_test = train_test_split(samples, labels, test_size=test_size, shuffle=shuffle, stratify=labels) | ||
| x_train, x_val, y_train, y_val = train_test_split(x_train, y_train, test_size=test_size, shuffle=shuffle, stratify=y_train) | ||
| classes = np.sort(np.unique(labels)) | ||
| y_train = to_categorical(y_train, num_classes=len(classes)) | ||
| y_test = to_categorical(y_test, num_classes=len(classes)) | ||
| y_val = to_categorical(y_val, num_classes=len(classes)) | ||
| return x_train, y_train, x_val, y_val, x_test, y_test | ||
|
|
||
| def build_model(shape, classes, hparams): | ||
| print(hparams) | ||
| N = shape[2] | ||
| F = shape[3] - N | ||
| frames = shape[1] | ||
|
|
||
| def get_feature_matrix(x, frame, N, F): | ||
| x = tf.slice(x, [0, frame, 0, N], [-1, 1, N, F]) | ||
| x = tf.squeeze(x, axis=[1]) | ||
| return x | ||
|
|
||
| def get_correlation_matrix(x, frame, N, F): | ||
| x = tf.slice(x, [0, frame, 0, 0], [-1, 1, N, N]) | ||
| x = tf.squeeze(x, axis=[1]) | ||
| return x | ||
|
|
||
| input_0 = tf.keras.Input((frames, N, F + N)) | ||
|
|
||
| gans = [] | ||
| for frame in range(frames): | ||
| feature_matrix = tf.keras.layers.Lambda( | ||
| get_feature_matrix, | ||
| arguments={'frame': frame, 'N': N, 'F': F} | ||
| )(input_0) | ||
|
|
||
| correlation_matrix = tf.keras.layers.Lambda( | ||
| get_correlation_matrix, | ||
| arguments={'frame': frame, 'N': N, 'F': F} | ||
| )(input_0) | ||
|
|
||
| x = sp.layers.GraphAttention(hparams['output_shape'])([feature_matrix, correlation_matrix]) | ||
| x = tf.keras.layers.Flatten()(x) | ||
| gans.append(x) | ||
|
|
||
| combine = tf.keras.layers.Concatenate()(gans) | ||
| reshape = tf.keras.layers.Reshape((frames, N * hparams['output_shape']))(combine) | ||
| lstm = tf.keras.layers.LSTM(hparams['hidden_units'])(reshape) | ||
| dropout = tf.keras.layers.Dropout(hparams['dropout'])(lstm) | ||
| out = tf.keras.layers.Dense(classes, activation='softmax')(dropout) | ||
|
|
||
| model = tf.keras.Model(inputs=[input_0], outputs=out) | ||
| model.compile( | ||
| optimizer=tf.keras.optimizers.Adam(learning_rate=hparams['learning_rate']), | ||
| loss='categorical_crossentropy', | ||
| metrics=[ | ||
| 'accuracy', | ||
| Recall(class_id=0, name='recall'), | ||
| Precision(class_id=0, name='precision'), | ||
| ] | ||
| ) | ||
| model.summary() | ||
| return model | ||
|
|
||
| def run_trial(path, step, model, hparams, x_train, y_train, x_val, y_val, x_test, y_test, epochs): | ||
| with tf.summary.create_file_writer(path).as_default(): | ||
| hp.hparams(hparams) | ||
| model.fit(x_train, y_train, epochs=epochs, batch_size=32, shuffle=True, validation_data=(x_val, y_val)) | ||
| loss, accuracy, recall, precision = model.evaluate(x_test, y_test) | ||
| tf.summary.scalar('accuracy', accuracy, step=step) | ||
| tf.summary.scalar('recall', recall, step=step) | ||
| tf.summary.scalar('precision', precision, step=step) | ||
|
|
||
| def hparams_combinations(hparams): | ||
| hp.hparams_config( | ||
| hparams=list(hparams.values()), | ||
| metrics=[ | ||
| hp.Metric('accuracy', display_name='Accuracy'), | ||
| hp.Metric('recall', display_name='Recall'), | ||
| hp.Metric('precision', display_name='Precision'), | ||
| ] | ||
| ) | ||
| hparams_keys = list(hparams.keys()) | ||
| hparams_values = list(product(*[ | ||
| h.domain.values | ||
| for h in hparams.values() | ||
| ])) | ||
| hparams = [ | ||
| dict(zip(hparams_keys, values)) | ||
| for values in hparams_values | ||
| ] | ||
| shuffle(hparams) | ||
| return hparams | ||
|
|
||
| def tune_model(dataset_name, data): | ||
| LOGS_DIR = join('./logs/generic', dataset_name) | ||
| os.makedirs(LOGS_DIR, exist_ok=True) | ||
| # Prepare the data | ||
| x_train, y_train, x_val, y_val, x_test, y_test = adapt_data(data) | ||
| # Set tuning session | ||
| counter = 0 | ||
| # Parameters to be tuned | ||
| hparams = { | ||
| 'learning_rate': [1e-4, 5e-4, 1e-3], | ||
| 'hidden_units': [8, 16, 32, 64], | ||
| 'output_shape': [8, 16, 32, 64], | ||
| 'dropout': [0.00, 0.05, 0.10, 0.15, 0.20], | ||
| } | ||
| hparams = { | ||
| key: hp.HParam(key, hp.Discrete(value)) | ||
| for key, value in hparams.items() | ||
| } | ||
| hparams = hparams_combinations(hparams) | ||
| for hparam in hparams: | ||
| # Build the model | ||
| model = build_model(data['data'].shape, len(data['labels_encoder']), hparam) | ||
| # Run session | ||
| run_name = f'run-{counter}' | ||
| print(f'--- Starting trial: {run_name}') | ||
| print(hparam) | ||
| run_trial( | ||
| join(LOGS_DIR, run_name), | ||
| counter, | ||
| model, | ||
| hparam, | ||
| x_train, | ||
| y_train, | ||
| x_val, | ||
| y_val, | ||
| x_test, | ||
| y_test, | ||
| epochs=50 | ||
| ) | ||
| counter += 1 | ||
|
|
||
|
|
||
| if __name__ == '__main__': | ||
| dataset = {} | ||
|
|
||
| dataset['tuh_eeg_abnormal'] = TUHEEGAbnormalDataset('../../data/tuh_eeg_abnormal/v2.0.0/edf') | ||
|
|
||
| dataset['tuh_eeg_artifact'] = TUHEEGArtifactDataset('../../data/tuh_eeg_artifact/v1.0.0/edf') | ||
| dataset['tuh_eeg_artifact'].set_minimum_event_duration(4) | ||
|
|
||
| dataset['tuh_eeg_seizure'] = TUHEEGSeizureDataset('../../data/tuh_eeg_seizure/v1.5.2/edf') | ||
| dataset['tuh_eeg_seizure'].set_minimum_event_duration(4) | ||
|
|
||
| # dataset['eegmmidb'] = EEGMMIDBDataset('../../data/physionet.org/files/eegmmidb/1.0.0') | ||
| # dataset['eegmmidb'].set_minimum_event_duration(4) | ||
|
|
||
| dataset['chbmit'] = CHBMITDataset('../../data/physionet.org/files/chbmit/1.0.0') | ||
| dataset['chbmit'].set_minimum_event_duration(4) | ||
|
|
||
| """ | ||
| Note: You can just use paths as values in the dictionary | ||
| and comment-out the first line of the following for cycle ;) | ||
| """ | ||
|
|
||
| for key, value in dataset.items(): | ||
| value = build_data(value) | ||
| tune_model(key, value) |
Oops, something went wrong.