model

[xiaonig]

您好请问有CNN-TCN主模型的代码吗

[TomMakesThings]

您好, 虽然我做三个模型在'ECG_Classifier.ipynb', 但是最好的模型是'TimeSeriesClassifier'。看看'Train Arrhythmia Classifier'下面的代码。特别是:

class TimeSeriesClassifier(BaseClassifier):
  def __init__(self, train_samples, train_labels, validation_samples=[], validation_labels=[], ...

...

# Set the model's parameters
model_kwargs = {'learning_rate': 0.001, 'dropout_rate': 0.3, 'architecture': "CNN",
                'tcn': True, 'attention': False, 'batch_norm': True, 'amsgrad': False}

...

arrhythmia_model_data = {}
...
# Otherwise set as the original training data with the arrhythmia class labels
arrhythmia_model_data['train_samples'] = X_train
arrhythmia_model_data['train_labels'] = y_train_new
...
# Otherwise create an arrhythmia classifier without transferring layers
arrhythmia_model = TimeSeriesClassifier(**arrhythmia_model_data, **model_kwargs)
...
 # Train the new model
 arrhythmia_model.train(epochs=200, batch_size=32, save_best=True, file_name=arrhythmia_model_file,
                      log_dir='logs/arrhythmia_fit/')

我两年前写了代码, 希望这回答您的问题!