def initialize_ae_diagnostics(self, n_epochs):
"""
initialize diagnostic variables for autoencoder network.
"""
self.train_time = 0
self.test_time = 0
self.best_weight_dict = None
# data-dependent diagnostics for train, validation, and test set
self.diag['train'] = NNetDataDiag(n_epochs=n_epochs, n_samples=self.data.n_train)
self.diag['val'] = NNetDataDiag(n_epochs=n_epochs, n_samples=self.data.n_val)
self.diag['test'] = NNetDataDiag(n_epochs=n_epochs, n_samples=self.data.n_test)
# network parameters
self.diag['network'] = {}
self.diag['network']['l2_penalty'] = np.zeros(n_epochs, dtype=Cfg.floatX)
def initialize_diagnostics(self, n_epochs):
"""
initialize diagnostics for the neural network
"""
# data-dependent diagnostics for train, validation, and test set
self.diag['train'] = NNetDataDiag(n_epochs=n_epochs, n_samples=self.data.n_train)
self.diag['val'] = NNetDataDiag(n_epochs=n_epochs, n_samples=self.data.n_val)
self.diag['test'] = NNetDataDiag(n_epochs=n_epochs, n_samples=self.data.n_test)
# network parameter diagnostics
self.diag['network'] = NNetParamDiag(self, n_epochs=n_epochs)
# Best results (highest AUC on test set)
self.auc_best = 0
self.auc_best_epoch = 0 # determined by highest AUC on test set
self.best_weight_dict = None