def _train_svi(self, train_loader, epochs, lr, device):
self.device = device
print("\n == SVI training ==")
optimizer = pyro.optim.Adam({"lr": lr})
elbo = TraceMeanField_ELBO()
svi = SVI(self.model, self.guide, optimizer, loss=elbo)
loss_list = []
accuracy_list = []
start = time.time()
for epoch in range(epochs):
loss = 0.0
correct_predictions = 0.0
for x_batch, y_batch in train_loader:
x_batch = x_batch.to(device)
y_batch = y_batch.to(device)
labels = y_batch.argmax(-1)
loss += svi.step(x_data=x_batch, y_data=labels)
outputs = self.forward(x_batch, n_samples=10)
predictions = outputs.argmax(dim=-1)
correct_predictions += (predictions == labels).sum().item()
if DEBUG:
print("\n",
pyro.get_param_store()["model.0.weight_loc"][0][:5])
print("\n", predictions[:10], "\n", labels[:10])
total_loss = loss / len(train_loader.dataset)
accuracy = 100 * correct_predictions / len(train_loader.dataset)
print(
f"\n[Epoch {epoch + 1}]\t loss: {total_loss:.2f} \t accuracy: {accuracy:.2f}",
end="\t")
loss_list.append(loss)
accuracy_list.append(accuracy)
execution_time(start=start, end=time.time())
self.save()
plot_loss_accuracy(dict={
'loss': loss_list,
'accuracy': accuracy_list
},
path=TESTS + self.name + "/" + self.name +
"_training.png")
mode='auto')
save_weights = ModelCheckpoint('model.h5',
monitor='val_loss',
save_best_only=True)
history = model.fit_generator(train_generator,
steps_per_epoch=nb_train_samples // batch_size,
epochs=epochs,
validation_data=validation_generator,
validation_steps=nb_validation_samples //
batch_size,
callbacks=[save_weights, early_stopping])
historydf = pd.DataFrame(history.history, index=history.epoch)
utils.plot_loss_accuracy(history)
test_datagen = ImageDataGenerator(rescale=1. / 255)
test_generator = test_datagen.flow_from_directory(test_dir,
target_size=(img_width,
img_height),
batch_size=batch_size,
class_mode=None,
shuffle=False)
#print(test_generator.filenames)
probabilities = model.predict_generator(test_generator,
nb_test_samples // batch_size)
mapper = {}
i = 0
for file in test_generator.filenames:
import matplotlib.pyplot as plt
import pandas as pd
from sklearn.model_selection import train_test_split
import utils
X, y = make_classification(n_samples=1000, n_features=2, n_redundant=0, \
n_informative=2, random_state=0, n_clusters_per_class=1)
print(X.shape)
X_train, X_validation, y_train, y_validation = train_test_split(X, y, test_size=0.33, random_state=100)
print(X_train.shape)
print(X_validation.shape)
utils.plot_data(X_train, y_train)
#perceptron model for binary classification
model = Sequential()
model.add(Dense(units=1, input_shape=(2,), activation='sigmoid'))
model.compile(optimizer='sgd', loss='mean_squared_error', metrics=['accuracy'])
history = model.fit(x=X_train, y=y_train, verbose=3, epochs=1000, validation_data=(X_validation,y_validation), batch_size=10)
print(model.summary())
print(model.get_weights())
historydf = pd.DataFrame(history.history, index=history.epoch)
utils.plot_loss_accuracy(history)
y_pred = model.predict_classes(X, verbose=0)
