X_train, Y_train = generate_linear(n=100)
X_test, Y_test = generate_linear(n=100)
elif args.dataset == 'xor':
X_train, Y_train = generate_XOR_easy()
X_test, Y_test = generate_XOR_easy()
else:
raise RuntimeError('Dataset Not Found')
net = Net()
if args.criterion == 'mse':
criterion = nn.MSE()
elif args.criterion == 'crossentropy':
criterion = nn.CrossEntropy()
else:
raise RuntimeError('Criterion Not Found')
if args.optimizer == 'sgd':
optimizer = optim.SGD(net.parameters(), lr=args.lr, momentum=args.momentum)
elif args.optimizer == 'adagrad':
optimizer = optim.Adagrad(net.parameters(), lr=args.lr)
else:
raise RuntimeError('Optimizer Not Found')
model = Model(net, criterion, optimizer)
train_history = model.train(X_train, Y_train, epochs=args.epochs)
test_history = model.test(X_test, Y_test)
show_history(train_history)
show_result(X_test, Y_test, test_history['predict'])
# mlp.evaluate_training()
# evaluate_file = input("What is the name of the file to evaluate the model on?\n")
# mlp.evaluate(evaluate_file)
#
# save_weights = input("Do you want to save weights of the model? (yes | no)\n").lower()
# while save_weights != "yes" and save_weights != "no":
# save_weights = input("Do you want to save weights of the model? (yes | no)\n").lower()
# if save_weights == "yes":
# save_weights_file = input("Please provide file name where the weights should be saved to:\n")
# mlp.save_weights(save_weights_file)
# UNCOMMENT WHEN YOU WANT TO FINALLY TEST THE MODEL
# First round
print("First round:")
training_history = mlp.train()
show_history(training_history)
training_error_first_round = mlp.evaluate_training()
evaluate_file = input(
"What is the name of the file to evaluate the model on?\n")
testing_error_first_round = mlp.evaluate(evaluate_file)
first_round_error = (training_error_first_round *
0.3) + (testing_error_first_round * 0.7)
print(
"Total error in first round (training (30%), testing (70%)) was: {}\n".
format(first_round_error))
# Second round
print("Second round:")
load_weights_file = input(
"What is the name of the file which you want to load weights from?\n")
mlp.load_weights(load_weights_file)
if args.load:
net = torch.load(args.load)
optimizer = optim.SGD(net.parameters(),
lr=args.lr,
momentum=0.9,
weight_decay=5e-4)
criterion = nn.CrossEntropyLoss()
model = Model(net, optimizer, criterion)
if args.trainable:
train_history = model.train(train_loader,
epochs=args.epochs,
val_loader=test_loader)
show_history(train_history,
f'history_{args.net}_{args.pretrained}.jpg')
with open(f'score_{args.net}_{args.pretrained}.txt',
'w') as score_file:
print(train_history['accuracy'], file=score_file)
print(train_history['val_accuracy'], file=score_file)
if args.save:
model.save(args.save)
test_history = model.test(test_loader)
show_confusion_matrix(
test_history['truth'], test_history['predict'],
f'confusion_matrix_{args.net}_{args.pretrained}.jpg')
if train:
model = create_model_2_lenet()
model.summary()
opt = Adam()
model.compile(loss=categorical_crossentropy,
optimizer=opt,
metrics=['accuracy'])
start = time()
history_object = model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
validation_data=(x_test, y_test),
shuffle=True,
callbacks=[checkpoint])
print("Training time:", time() - start)
show_history(history_object)
else:
model = load_model(os.path.join(dir_save, model_name))
model.summary()
print("H5 Output: " + str(model.output.op.name))
print("H5 Input: " + str(model.input.op.name))
# Score trained model.
scores = model.evaluate(x_test, y_test, verbose=1)
print('Validation loss:', scores[0])
print('Validation accuracy:', scores[1])
samples, targets = next(val_gen)
preds = samples[:, -1, 1]
mae = np.mean(np.abs(preds - targets))
batch_maes.append(mae)
print(np.mean(batch_maes))
# 开始使用基准评估
evaluate_naive_method()
# 训练并评估一个一维卷积神经网络
model = Sequential()
model.add(
layers.Conv1D(32,
5,
activation='relu',
input_shape=(None, float_data.shape[-1])))
model.add(layers.MaxPooling1D(3))
model.add(layers.Conv1D(32, 5, activation='relu'))
model.add(layers.MaxPooling1D(3))
model.add(layers.Conv1D(32, 5, activation='relu'))
model.add(layers.GlobalMaxPooling1D())
model.add(layers.Dense(1))
model.compile(optimizer=RMSprop(), loss='mae')
history = model.fit_generator(train_gen,
steps_per_epoch=500,
epochs=20,
validation_data=val_gen,
validation_steps=val_steps)
show_history(history)