def test_simple_mlp():
Asamples = np.random.multivariate_normal([6, 6], [[1, 0], [0, 1]], 200)
Bsamples = np.random.multivariate_normal([1, 1], [[1, 0], [0, 1]], 200)
x_train = np.vstack((Asamples, Bsamples))
y_train = np.vstack((np.array([[0, 1]] * 200), np.array([[1, 0]] * 200)))
print(x_train.shape, y_train.shape)
num_classes = 2
batch_size = 10
epochs = 10
model = Sequential()
model.add(Dense(units=2, activation='relu', input_shape=(2, )))
model.add(Dense(units=num_classes, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer='sgd',
learning_rate=0.05)
history = model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(None, None))
def test_mlp_mnist():
train_set, test_set = mnist(one_hot=True)
x_train, y_train = train_set[0], train_set[1]
x_test, y_test = test_set[0], test_set[1]
x_train = x_train.reshape(x_train.shape[0],
x_train.shape[1] * x_train.shape[2])
x_test = x_test.reshape(x_test.shape[0], x_test.shape[1] * x_test.shape[2])
num_classes = 10
batch_size = 32
epochs = 1
model = Sequential()
model.add(Dense(units=256, activation='relu', input_shape=(784, )))
model.add(Dense(units=128, activation='relu'))
model.add(Dense(units=64, activation='relu'))
model.add(Dense(num_classes, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer='momentum',
learning_rate=0.05)
history = model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(x_test, y_test))
score = model.evaluate(x_test, y_test, verbose=0)
def test_mlp():
model = Sequential()
model.add(Dense(units=512, activation='relu', input_shape=(784, )))
model.add(Dense(units=512, activation='relu'))
model.add(Dense(num_classes, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy', optimizer='RMSprop')
history = model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(None, None))
score = model.evaluate(x_test, y_test, verbose=0)
def test_mlp_forward():
print("\n")
x = np.array([[1, 0], [0, 1], [1, 1], [0, 0]])
y = np.array([[1, 0], [1, 0], [0, 1], [0, 1]])
print("x shape:", x.shape, "y shape:", y.shape)
print("\n")
print("x", x)
print("y", y)
print("\n")
num_classes = 2
batch_size = 1
epochs = 20
# Build network
input0 = InputLayer(shape=[2])
dense1 = Dense(units=2, activation='relu', input_shape=[2])
input0.outbound.append(dense1)
dense1.inbound.append(input0)
# Init parameters of dense1
dense1.name = 'dense1'
kernel = np.array([[1., 2.], [2., 1.]])
bias = np.array([1., 1.])
dense1.params = {'kernel': kernel, 'bias': bias}
dense1.isInitialized = True
print("kernel", kernel)
print("bias", bias)
# ForwardPropagation
inputs = input0.ForwardPropagation(feed=x)
print("inputs", inputs)
f1 = dense1.ForwardPropagation()
print("f1", f1)
f1_expect = np.matmul(inputs, kernel) + bias
print("expect f1", f1_expect)
def test_mlp_backward():
print("\n")
print("\n")
x = np.array([[1, 0], [1, 1], [0, 1]])
y = np.array([[0., 1.], [1., 0.], [0, 1]])
print("x shape:", x.shape, "y shape:", y.shape)
print("\n")
print("x", x)
print("y", y)
print("\n")
# Build network
input0 = InputLayer(shape=[2])
dense1 = Dense(units=2, activation='linear', input_shape=(2, ))
loss2 = SoftmaxCategoricalCrossentropy()
input0.outbound.append(dense1)
dense1.inbound.append(input0)
dense1.outbound.append(loss2)
loss2.inbound.append(dense1)
# Init parameters of dense1
dense1.name = 'dense1'
kernel = np.array([[1., 2.], [2., 1.]])
bias = np.array([1., 1.])
dense1.params = {'kernel': kernel, 'bias': bias}
dense1.grads = {'kernel': np.array([]), 'bias': np.array([])}
dense1.isInitialized = True
print("kernel", kernel)
print("bias", bias)
# ForwardPropagation
inputs = input0.ForwardPropagation(feed=x)
print("inputs", inputs)
f1 = dense1.ForwardPropagation()
print("f1", f1)
# BackwardPropagation
logits = np.array([[-10., 10.], [10., -10.], [-10., 10.]])
loss2.y_true = np.array([[0., 1.], [1., 0.], [0., 1.]])
loss2.input_value = logits
loss2.output_value = softmax_categorical_crossentropy(y_true=loss2.y_true,
logits=logits)
print("loss2.output_value", loss2.output_value)
grad_loss = loss2.BackwardPropagation(grads=None)
print("grad_loss", grad_loss)
grad_dense = dense1.BackwardPropagation(grad=grad_loss)
print("grad_dense", grad_dense)
def test_single_layer():
Asamples = np.random.multivariate_normal([6, 6], [[1, 0], [0, 1]], 200)
Bsamples = np.random.multivariate_normal([1, 1], [[1, 0], [0, 1]], 200)
Csamples = np.random.multivariate_normal([12, 12], [[1, 0], [0, 1]], 200)
# plt.figure()
# plt.plot(Asamples[:, 0], Asamples[:, 1], 'r.')
# plt.plot(Bsamples[:, 0], Bsamples[:, 1], 'b.')
# plt.plot(Csamples[:, 0], Csamples[:, 1], 'g.')
# plt.show()
x_train = np.vstack((Asamples, Bsamples, Csamples))
y_train = np.vstack(
(np.array([[0, 0, 1]] * 200), np.array([[0, 1, 0]] * 200),
np.array([[1, 0, 0]] * 200)))
print(x_train.shape, y_train.shape)
num_classes = 3
batch_size = 20
epochs = 100
model = Sequential()
model.add(Dense(units=num_classes, activation='softmax',
input_shape=(2, )))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer='sgd',
learning_rate=0.005)
history = model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(None, None))
from SemiFlow.layer import Dense
from SemiFlow.Model import Sequential
from SemiFlow.utils.dataset import mnist
train_set, valid_set, test_set = mnist(one_hot=True)
x_train, y_train = train_set[0], train_set[1]
x_test, y_test = test_set[0], test_set[1]
x_val, y_val = valid_set[0], valid_set[1]
num_classes = 10
batch_size = 32
epochs = 3
model = Sequential()
model.add(Dense(units=256, activation='relu', input_shape=(784, )))
model.add(Dense(units=128, activation='relu'))
model.add(Dense(units=64, activation='relu'))
model.add(Dense(num_classes, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer='RMSprop',
learning_rate=0.05,
metrics=['train_loss', 'val_loss'])
history = model.fit(
x_train,
y_train,
batch_size=batch_size,
num_classes = 10
batch_size = 32
epochs = 1
model = Sequential()
model.add(
Conv2D(32,
kernel_size=(3, 3),
activation='relu',
input_shape=(28, 28, 1),
dtype='float32'))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pooling_size=(3, 3)))
model.add(Flatten())
model.add(Dense(units=64, activation='relu'))
model.add(Dense(num_classes, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer='RMSprop',
learning_rate=0.05)
history = model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(x_val, y_val))
score = model.evaluate(x_test, y_test, verbose=0)