def get_seq_model():
"""Define three channel input shape depending on image data format."""
if K.image_data_format() == 'channels_first':
input_shape = (3, img_width, img_height)
else:
input_shape = (img_width, img_height, 3)
# Initialize CNN by creating a sequential model.
model = Sequential()
model.add(Conv2D(32, (3, 3), input_shape=input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Flatten())
model.add(Dense(64))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(2))
model.add(Activation('sigmoid'))
model.compile(
loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
return model
def inner(x):
x = LayerNormalization()(x)
x = Activation("relu")(x)
x = Convolution2D(channels, 3, strides=strides, **params)(x)
x = Dropout(drop_rate)(x) if drop_rate > 0 else x
x = LayerNormalization()(x)
x = Activation("relu")(x)
x = Convolution2D(channels, 3, **params)(x)
return x
def residual_block(
inputs,
num_filters=16,
kernel_size=3,
strides=1,
activation="relu",
batch_normalization=True,
conv_first=True,
):
"""2D Convolution-Batch Normalization-Activation stack builder
# Arguments
inputs (tensor): input tensor from input image or previous layer
num_filters (int): Conv2D number of filters
kernel_size (int): Conv2D square kernel dimensions
strides (int): Conv2D square stride dimensions
activation (string): activation name
batch_normalization (bool): whether to include batch normalization
conv_first (bool): conv-bn-activation (True) or
bn-activation-conv (False)
# Returns
x (tensor): tensor as input to the next layer
"""
conv = Conv2D(
num_filters,
kernel_size=kernel_size,
strides=strides,
padding="same",
kernel_initializer="he_normal",
kernel_regularizer=l2(1e-4),
activation=None,
)
conv2 = Conv2D(
num_filters,
kernel_size=kernel_size,
strides=strides,
padding="same",
kernel_initializer="he_normal",
kernel_regularizer=l2(1e-4),
activation="linear",
)
x = conv(inputs)
x = BatchNormalization()(x)
x = Activation(activation)(x)
x = conv2(x)
x = add([inputs, x])
x = BatchNormalization()(x)
x = Activation(activation)(x)
return x
def DNNclassifier_crps(self, p, num_cut, optimizer, seeding):
tf.set_random_seed(seeding)
inputs = Input(shape=(p,))
if isinstance(optimizer, str):
opt = optimizer
else:
opt_name = optimizer.__class__.__name__
opt_config = optimizer.get_config()
opt_class = getattr(optimizers, opt_name)
opt = opt_class(**opt_config)
for i, n_neuron in enumerate(self.hidden_list):
if i == 0:
net = Dense(n_neuron, kernel_initializer = 'he_uniform')(inputs)
else:
net = Dense(n_neuron, kernel_initializer = 'he_uniform')(net)
net = Activation(activation = 'elu')(net)
net = BatchNormalization()(net)
net = Dropout(rate=self.dropout_list[i])(net)
softmaxlayer = Dense(num_cut + 1, activation='softmax',
kernel_initializer = 'he_uniform')(net)
output = Lambda(self.tf_cumsum)(softmaxlayer)
model = Model(inputs = [inputs], outputs=[output])
model.compile(optimizer=opt, loss=self.crps_loss)
return model
def build_elu_cnn(input_shape, output_size):
"""Build a variation of the CNN implemented in the ELU paper.
https://arxiv.org/abs/1511.07289
"""
def layers(n, channels, kernel):
return sum(([
Convolution2D(channels, kernel_size=kernel, padding="same"),
ELU()
] for i in range(n)), [])
model = Sequential(
[
Convolution2D(
384, kernel_size=3, padding="same", input_shape=input_shape)
] + layers(1, 384, 3) + [MaxPooling2D(pool_size=(2, 2))] +
layers(1, 384, 1) + layers(1, 384, 2) + layers(2, 640, 2) +
[MaxPooling2D(pool_size=(2, 2))] + layers(1, 640, 1) +
layers(3, 768, 2) + [MaxPooling2D(pool_size=(2, 2))] +
layers(1, 768, 1) + layers(2, 896, 2) +
[MaxPooling2D(pool_size=(2, 2))] + layers(1, 896, 3) +
layers(2, 1024, 2) + [
MaxPooling2D(pool_size=(2, 2)),
Convolution2D(output_size, kernel_size=1, padding="same"),
GlobalAveragePooling2D(),
Activation("softmax")
])
model.compile(optimizer=SGD(momentum=0.9),
loss="categorical_crossentropy",
metrics=["accuracy"])
return model
def ensure_softmax_output(model):
"""
Adds a softmax layer on top of the logits layer, in case the output layer
is a logits layer.
Parameters
----------
model : Keras Model
The original model
Returns
-------
new_model : Keras Model
The modified model
"""
if 'softmax' not in model.output_names:
if 'logits' in model.output_names:
output = Activation('softmax', name='softmax')(model.output)
new_model = Model(inputs=model.input, outputs=output)
else:
raise ValueError('The output layer is neither softmax nor logits')
else:
new_model = model
return new_model
def NN_huaweiv1(maxlen, embedding_matrix=None, class_num1=17, class_num2=12):
emb_layer = Embedding(
embedding_matrix.shape[0],
embedding_matrix.shape[1],
input_length=maxlen,
weights=[embedding_matrix],
trainable=False,
)
seq1 = Input(shape=(maxlen, ))
emb = emb_layer(seq1)
sdrop = SpatialDropout1D(rate=0.2)
lstm_layer = Bidirectional(CuDNNGRU(128, return_sequences=True))
gru_layer = Bidirectional(CuDNNGRU(128, return_sequences=True))
cnn1d_layer = Conv1D(64,
kernel_size=3,
padding="same",
kernel_initializer="he_uniform")
sd = sdrop(emb)
lstm1 = lstm_layer(sd)
gru1 = gru_layer(lstm1)
cnn1 = cnn1d_layer(gru1)
gru1 = concatenate([lstm1, gru1, cnn1])
att_1 = Attention(maxlen)(gru1)
att_2 = Attention(maxlen)(gru1)
att_3 = Attention(maxlen)(gru1)
att_4 = Attention(maxlen)(gru1)
x1 = Activation(activation="relu")(BatchNormalization()(Dense(128)(att_1)))
x2 = Activation(activation="relu")(BatchNormalization()(Dense(128)(att_2)))
x3 = Activation(activation="relu")(BatchNormalization()(Dense(128)(att_3)))
x4 = Activation(activation="relu")(BatchNormalization()(Dense(128)(att_4)))
pred1_1 = Dense(class_num1 - 10, activation='sigmoid')(x1)
pred1_2 = Dense(10, activation='sigmoid')(x2)
pred1 = concatenate([pred1_1, pred1_2], axis=-1, name='pred1')
pred2_1 = Dense(class_num2 - 9, activation='sigmoid')(x3)
pred2_2 = Dense(9, activation='sigmoid')(x4)
pred2 = concatenate(
[pred2_1, pred2_2], axis=-1, name='pred2'
) # Dense(class_num2, activation='sigmoid',name='pred2')(y)
model = Model(inputs=seq1, outputs=[pred1, pred2])
return model
def NN_huaweiv1(maxlen, embedding_matrix=None, class_num1=17, class_num2=12):
emb_layer = Embedding(
embedding_matrix.shape[0],
embedding_matrix.shape[1],
input_length=maxlen,
weights=[embedding_matrix],
trainable=False,
)
seq1 = Input(shape=(maxlen, ))
x1 = emb_layer(seq1)
sdrop = SpatialDropout1D(rate=0.2)
lstm_layer = Bidirectional(CuDNNGRU(128, return_sequences=True))
gru_layer = Bidirectional(CuDNNGRU(128, return_sequences=True))
cnn1d_layer = Conv1D(64,
kernel_size=3,
padding="same",
kernel_initializer="he_uniform")
x1 = sdrop(x1)
lstm1 = lstm_layer(x1)
gru1 = gru_layer(lstm1)
att_1 = Attention(maxlen)(lstm1)
att_2 = Attention(maxlen)(gru1)
cnn1 = cnn1d_layer(lstm1)
avg_pool = GlobalAveragePooling1D()
max_pool = GlobalMaxPooling1D()
x1 = concatenate([
att_1, att_2,
Attention(maxlen)(cnn1),
avg_pool(cnn1),
max_pool(cnn1)
])
x = Dropout(0.2)(Activation(activation="relu")(BatchNormalization()(
Dense(128)(x1))))
x = Activation(activation="relu")(BatchNormalization()(Dense(64)(x)))
pred1 = Dense(class_num1, activation='sigmoid', name='pred1')(x)
y = concatenate([x1, x])
y = Activation(activation="relu")(BatchNormalization()(Dense(64)(x)))
pred2 = Dense(class_num2, activation='sigmoid', name='pred2')(y)
model = Model(inputs=seq1, outputs=[pred1, pred2])
return model
def build_small_cnn(input_shape, output_size):
model = Sequential([
# conv1_*
Convolution2D(32,
kernel_size=3,
padding="same",
input_shape=input_shape),
Activation("relu"),
Convolution2D(32, kernel_size=3, padding="same"),
Activation("relu"),
MaxPooling2D(pool_size=(2, 2)),
# conv2_*
Convolution2D(64, kernel_size=3, padding="same"),
Activation("relu"),
Convolution2D(64, kernel_size=3, padding="same"),
Activation("relu"),
MaxPooling2D(pool_size=(2, 2)),
# Fully connected
Flatten(),
Dense(512),
Activation("relu"),
Dense(512),
Activation("relu"),
Dense(output_size),
Activation("softmax")
])
model.compile(loss="categorical_crossentropy",
optimizer="adam",
metrics=["accuracy"])
return model
def build_cnn(input_shape, output_size):
kwargs = {"kernel_size": 3, "activation": "relu", "padding": "same"}
model = Sequential([
# conv1_*
Convolution2D(64, input_shape=input_shape, **kwargs),
BatchRenormalization(),
Convolution2D(64, **kwargs),
BatchRenormalization(),
MaxPooling2D(pool_size=(2, 2)),
Dropout(0.25),
# conv2_*
Convolution2D(128, **kwargs),
BatchRenormalization(),
Convolution2D(128, **kwargs),
BatchRenormalization(),
MaxPooling2D(pool_size=(2, 2)),
Dropout(0.25),
# conv3_*
Convolution2D(256, **kwargs),
BatchRenormalization(),
Convolution2D(256, **kwargs),
BatchRenormalization(),
MaxPooling2D(pool_size=(2, 2)),
Dropout(0.25),
# Fully connected
Flatten(),
Dense(1024),
Activation("relu"),
Dropout(0.5),
Dense(512),
Activation("relu"),
Dropout(0.5),
Dense(output_size),
Activation("softmax")
])
model.compile(loss="categorical_crossentropy",
optimizer="adam",
metrics=["accuracy"])
return model
def build_lr(input_shape, output_size):
model = Sequential([
Flatten(input_shape=input_shape),
Dense(output_size),
Activation("softmax")
])
model.compile(loss="categorical_crossentropy",
optimizer="adam",
metrics=["accuracy"])
return model
def face_impl(input_shape, output_size):
x = Input(shape=input_shape)
e = modelf(input_shape, embedding)(x)
y = Dense(output_size)(e)
y = Activation("softmax")(y)
model = Model(x, y)
model.compile("adam",
"sparse_categorical_crossentropy",
metrics=["accuracy"])
return model
def build_lstm_model(input_data, output_size, neurons=20, activ_func='linear',
dropout=0.25, loss='mae', optimizer='adam'):
model = Sequential()
model.add(CuDNNLSTM(neurons, input_shape=(input_data.shape[1], input_data.shape[2]), return_sequences=True))
model.add(Dropout(dropout))
model.add(CuDNNLSTM(neurons, input_shape=(input_data.shape[1], input_data.shape[2])))
model.add(Dropout(dropout))
model.add(Dense(units=output_size))
model.add(Activation(activ_func))
model.compile(loss=loss, optimizer=optimizer)
return model
def inception_block_1a(X):
"""
Implementation of an inception block
"""
X_3x3 = Conv2D(96, (1, 1), data_format='channels_first', name='inception_3a_3x3_conv1')(X)
X_3x3 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3a_3x3_bn1')(X_3x3)
X_3x3 = Activation('relu')(X_3x3)
X_3x3 = ZeroPadding2D(padding=(1, 1), data_format='channels_first')(X_3x3)
X_3x3 = Conv2D(128, (3, 3), data_format='channels_first', name='inception_3a_3x3_conv2')(X_3x3)
X_3x3 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3a_3x3_bn2')(X_3x3)
X_3x3 = Activation('relu')(X_3x3)
X_5x5 = Conv2D(16, (1, 1), data_format='channels_first', name='inception_3a_5x5_conv1')(X)
X_5x5 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3a_5x5_bn1')(X_5x5)
X_5x5 = Activation('relu')(X_5x5)
X_5x5 = ZeroPadding2D(padding=(2, 2), data_format='channels_first')(X_5x5)
X_5x5 = Conv2D(32, (5, 5), data_format='channels_first', name='inception_3a_5x5_conv2')(X_5x5)
X_5x5 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3a_5x5_bn2')(X_5x5)
X_5x5 = Activation('relu')(X_5x5)
X_pool = MaxPooling2D(pool_size=3, strides=2, data_format='channels_first')(X)
X_pool = Conv2D(32, (1, 1), data_format='channels_first', name='inception_3a_pool_conv')(X_pool)
X_pool = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3a_pool_bn')(X_pool)
X_pool = Activation('relu')(X_pool)
X_pool = ZeroPadding2D(padding=((3, 4), (3, 4)), data_format='channels_first')(X_pool)
X_1x1 = Conv2D(64, (1, 1), data_format='channels_first', name='inception_3a_1x1_conv')(X)
X_1x1 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3a_1x1_bn')(X_1x1)
X_1x1 = Activation('relu')(X_1x1)
# CONCAT
inception = concatenate([X_3x3, X_5x5, X_pool, X_1x1], axis=1)
return inception
def inception_block_1b(X):
X_3x3 = Conv2D(96, (1, 1), data_format='channels_first', name='inception_3b_3x3_conv1')(X)
X_3x3 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3b_3x3_bn1')(X_3x3)
X_3x3 = Activation('relu')(X_3x3)
X_3x3 = ZeroPadding2D(padding=(1, 1), data_format='channels_first')(X_3x3)
X_3x3 = Conv2D(128, (3, 3), data_format='channels_first', name='inception_3b_3x3_conv2')(X_3x3)
X_3x3 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3b_3x3_bn2')(X_3x3)
X_3x3 = Activation('relu')(X_3x3)
X_5x5 = Conv2D(32, (1, 1), data_format='channels_first', name='inception_3b_5x5_conv1')(X)
X_5x5 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3b_5x5_bn1')(X_5x5)
X_5x5 = Activation('relu')(X_5x5)
X_5x5 = ZeroPadding2D(padding=(2, 2), data_format='channels_first')(X_5x5)
X_5x5 = Conv2D(64, (5, 5), data_format='channels_first', name='inception_3b_5x5_conv2')(X_5x5)
X_5x5 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3b_5x5_bn2')(X_5x5)
X_5x5 = Activation('relu')(X_5x5)
X_pool = AveragePooling2D(pool_size=(3, 3), strides=(3, 3), data_format='channels_first')(X)
X_pool = Conv2D(64, (1, 1), data_format='channels_first', name='inception_3b_pool_conv')(X_pool)
X_pool = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3b_pool_bn')(X_pool)
X_pool = Activation('relu')(X_pool)
X_pool = ZeroPadding2D(padding=(4, 4), data_format='channels_first')(X_pool)
X_1x1 = Conv2D(64, (1, 1), data_format='channels_first', name='inception_3b_1x1_conv')(X)
X_1x1 = BatchNormalization(axis=1, epsilon=0.00001, name='inception_3b_1x1_bn')(X_1x1)
X_1x1 = Activation('relu')(X_1x1)
inception = concatenate([X_3x3, X_5x5, X_pool, X_1x1], axis=1)
return inception
def build_lstm_timit(input_shape, output_size):
"""Build a simple LSTM to classify the phonemes in the TIMIT dataset"""
model = Sequential([
LSTM(256, unroll=True, input_shape=input_shape),
Dense(output_size),
Activation("softmax")
])
model.compile(optimizer="adam",
loss="sparse_categorical_crossentropy",
metrics=["accuracy"])
return model
def build_lstm_mnist(input_shape, output_size):
"""Build a small LSTM to recognize MNIST digits as permuted sequences"""
model = Sequential([
CuDNNLSTM(128, input_shape=input_shape),
Dense(output_size),
Activation("softmax")
])
model.compile(optimizer="adam",
loss="categorical_crossentropy",
metrics=["accuracy"])
return model
def build_lstm_lm(input_shape, output_size):
# LM datasets will report the vocab_size as output_size
vocab_size = output_size
model = Sequential([
Embedding(vocab_size + 1,
64,
mask_zero=True,
input_length=input_shape[0]),
LSTM(256, unroll=True, return_sequences=True),
LSTM(256, unroll=True),
Dense(output_size),
Activation("softmax")
])
model.compile(optimizer="adam",
loss="sparse_categorical_crossentropy",
metrics=["accuracy"])
return model
def build_all_conv_nn(input_shape, output_size):
"""Build a small variation of the best performing network from
'Springenberg, Jost Tobias, et al. "Striving for simplicity: The all
convolutional net." arXiv preprint arXiv:1412.6806 (2014)' which should
achieve approximately 91% in CIFAR-10.
"""
kwargs = {"activation": "relu", "border_mode": "same"}
model = Sequential([
# conv1
Convolution2D(96, 3, 3, input_shape=input_shape, **kwargs),
BatchRenormalization(),
Convolution2D(96, 3, 3, **kwargs),
BatchRenormalization(),
Convolution2D(96, 3, 3, subsample=(2, 2), **kwargs),
BatchRenormalization(),
Dropout(0.25),
# conv2
Convolution2D(192, 3, 3, **kwargs),
BatchRenormalization(),
Convolution2D(192, 3, 3, **kwargs),
BatchRenormalization(),
Convolution2D(192, 3, 3, subsample=(2, 2), **kwargs),
BatchRenormalization(),
Dropout(0.25),
# conv3
Convolution2D(192, 1, 1, **kwargs),
BatchRenormalization(),
Dropout(0.25),
Convolution2D(output_size, 1, 1, **kwargs),
GlobalAveragePooling2D(),
Activation("softmax")
])
model.compile(loss="categorical_crossentropy",
optimizer=SGD(momentum=0.9),
metrics=["accuracy"])
return model
if i % 15 == 0: return np.array([0, 0, 0, 1], dtype=np.float32)
elif i % 5 == 0: return np.array([0, 0, 1, 0], dtype=np.float32)
elif i % 3 == 0: return np.array([0, 1, 0, 0], dtype=np.float32)
else: return np.array([1, 0, 0, 0], dtype=np.float32)
def bin(i, num_digits):
return np.array([i >> d & 1 for d in range(num_digits)], dtype=np.float32)
NUM_DIGITS = 7
trX = np.array([bin(i, NUM_DIGITS) for i in range(1, 101)])
trY = np.array([fizzbuzz(i) for i in range(1, 101)])
model = Sequential()
model.add(Dense(64, input_dim=7))
model.add(Activation('tanh'))
model.add(Dense(4, input_dim=64))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
model.fit(trX, trY, epochs=3600, batch_size=64)
model.save('fizzbuzz_model.h5')
def representative_dataset_gen():
for i in range(100):
yield [trX[i:i + 1]]
converter = lite.TFLiteConverter.from_keras_model_file('fizzbuzz_model.h5')
def keras_build_fn(num_feature,
num_output,
is_sparse,
embedding_dim=-1,
num_hidden_layer=2,
hidden_layer_dim=512,
activation='elu',
learning_rate=1e-3,
dropout=0.5,
l1=0.0,
l2=0.0,
loss='categorical_crossentropy'):
"""Initializes and compiles a Keras DNN model using the Adam optimizer.
Args:
num_feature: number of features
num_output: number of outputs (targets, e.g., classes))
is_sparse: boolean whether input data is in sparse format
embedding_dim: int number of nodes in embedding layer; if value is <= 0 then
no embedding layer will be present in the model
num_hidden_layer: number of hidden layers
hidden_layer_dim: int number of nodes in the hidden layer(s)
activation: string
activation function for hidden layers; see https://keras.io/activations/
learning_rate: float learning rate for Adam
dropout: float proportion of nodes to dropout; values in [0, 1]
l1: float strength of L1 regularization on weights
l2: float strength of L2 regularization on weights
loss: string
loss function; see https://keras.io/losses/
Returns:
model: Keras.models.Model
compiled Keras model
"""
assert num_hidden_layer >= 1
inputs = Input(shape=(num_feature, ), sparse=is_sparse)
activation_func_args = ()
if activation.lower() == 'prelu':
activation_func = PReLU
elif activation.lower() == 'leakyrelu':
activation_func = LeakyReLU
elif activation.lower() == 'elu':
activation_func = ELU
elif activation.lower() == 'thresholdedrelu':
activation_func = ThresholdedReLU
else:
activation_func = Activation
activation_func_args = (activation)
if l1 > 0 and l2 > 0:
reg_init = lambda: regularizers.l1_l2(l1, l2)
elif l1 > 0:
reg_init = lambda: regularizers.l1(l1)
elif l2 > 0:
reg_init = lambda: regularizers.l2(l2)
else:
reg_init = lambda: None
if embedding_dim > 0:
# embedding layer
e = Dense(embedding_dim)(inputs)
x = Dense(hidden_layer_dim, kernel_regularizer=reg_init())(e)
x = activation_func(*activation_func_args)(x)
x = Dropout(dropout)(x)
else:
x = Dense(hidden_layer_dim, kernel_regularizer=reg_init())(inputs)
x = activation_func(*activation_func_args)(x)
x = Dropout(dropout)(x)
# add additional hidden layers
for _ in range(num_hidden_layer - 1):
x = Dense(hidden_layer_dim, kernel_regularizer=reg_init())(x)
x = activation_func(*activation_func_args)(x)
x = Dropout(dropout)(x)
x = Dense(num_output)(x)
preds = Activation('softmax')(x)
model = Model(inputs=inputs, outputs=preds)
model.compile(optimizer=Adam(lr=learning_rate), loss=loss)
return model
def InceptionModel(input_shape):
"""
Implementation of the Inception model used for FaceNet
Arguments:
input_shape -- shape of the images of the dataset
Returns:
model -- a Model() instance in Keras
"""
# Define the input as a tensor with shape input_shape
X_input = Input(input_shape)
# Zero-Padding
X = ZeroPadding2D((3, 3))(X_input)
# First Block
X = Conv2D(64, (7, 7), strides=(2, 2), name='conv1')(X)
X = BatchNormalization(axis=1, name='bn1')(X)
X = Activation('relu')(X)
# Zero-Padding + MAXPOOL
X = ZeroPadding2D((1, 1))(X)
X = MaxPooling2D((3, 3), strides=2)(X)
# Second Block
X = Conv2D(64, (1, 1), strides=(1, 1), name='conv2')(X)
X = BatchNormalization(axis=1, epsilon=0.00001, name='bn2')(X)
X = Activation('relu')(X)
# Zero-Padding + MAXPOOL
X = ZeroPadding2D((1, 1))(X)
# Second Block
X = Conv2D(192, (3, 3), strides=(1, 1), name='conv3')(X)
X = BatchNormalization(axis=1, epsilon=0.00001, name='bn3')(X)
X = Activation('relu')(X)
# Zero-Padding + MAXPOOL
X = ZeroPadding2D((1, 1))(X)
X = MaxPooling2D(pool_size=3, strides=2)(X)
# Inception 1: a/b/c
X = inception_block_1a(X)
X = inception_block_1b(X)
X = inception_block_1c(X)
# Inception 2: a/b
X = inception_block_2a(X)
X = inception_block_2b(X)
# Inception 3: a/b
X = inception_block_3a(X)
X = inception_block_3b(X)
# Top layer
X = AveragePooling2D(pool_size=(3, 3), strides=(1, 1), data_format='channels_first')(X)
X = Flatten()(X)
X = Dense(128, name='dense_layer')(X)
# L2 normalization
X = Lambda(lambda x: K.l2_normalize(x, axis=1))(X)
# Create model instance
model = Model(inputs=X_input, outputs=X, name='FaceRecoModel')
return model
dense_layers = [0, 1, 2]
layer_sizes = [4, 8, 16]
conv_layers = [1, 2]
for dense_layer in dense_layers:
for layer_size in layer_sizes:
for conv_layer in conv_layers:
NAME = f'Pneumonia-{IMG_SIZE}px-{NUM_SAMPLES}samples-{conv_layer}conv-{layer_size}nodes-{dense_layer}dense-{int(time.time())}'
tensorboard = TensorBoard(log_dir=f'logs/{NAME}')
print(NAME)
model = Sequential()
# format: Num of filters, window/step, dimensions
model.add(Conv2D(layer_size, (3, 3),
input_shape=x_train.shape[1:]))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(2, 2)))
print('Layer 0 generated')
for i in range(conv_layer - 1):
print(f'Layer {i + 1} generated.')
model.add(Conv2D(layer_size, (3, 3)))
model.add(Activation("relu"))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Flatten())
for l in range(dense_layer):
model.add(Dense(layer_size))
model.add(Activation("relu"))
model.add(Dense(1))
def create_alpha_zero_model(
depth,
input_shape,
policy_output_size,
num_filters=64,
activation="relu",
policy_factor=1.0,
):
input = tf.keras.Input(shape=input_shape, name="input")
conv = Conv2D(
num_filters,
kernel_size=3,
strides=1,
padding="same",
kernel_initializer="he_normal",
kernel_regularizer=l2(1e-4),
activation=None,
)
x = conv(input)
x = BatchNormalization()(x)
x = Activation(activation)(x)
block_output = residual_block(inputs=x, strides=1, num_filters=num_filters)
for _ in range(depth):
block_output = residual_block(inputs=block_output,
strides=1,
num_filters=num_filters)
# TODO: consider adding an extra conv layer here and for the policy head as
# well, see https://medium.com/oracledevs/lessons-from-alpha-zero-part-6-hyperparameter-tuning-b1cfcbe4ca9
value_conv_output = Conv2D(
num_filters // 2,
kernel_size=3,
strides=1,
padding="same",
kernel_initializer="he_normal",
kernel_regularizer=l2(1e-4),
activation=None,
)(block_output)
value_conv_output = BatchNormalization()(value_conv_output)
value_conv_output = Activation(activation)(value_conv_output)
value = Dense(
units=1,
kernel_regularizer=l2(1e-4),
kernel_initializer="he_normal",
activation="tanh",
name="value",
)(Flatten()(value_conv_output))
policy_conv_output = Conv2D(
num_filters // 2,
kernel_size=3,
strides=1,
padding="same",
kernel_initializer="he_normal",
kernel_regularizer=l2(1e-4),
activation=None,
)(block_output)
policy_conv_output = BatchNormalization()(policy_conv_output)
policy_conv_output = Activation(activation)(policy_conv_output)
policy = (Dense(
units=policy_output_size,
kernel_regularizer=l2(1e-4),
kernel_initializer="he_normal",
activation=None,
)(Flatten()(policy_conv_output)) * policy_factor)
policy = Activation("softmax", name="policy")(policy)
# policy = tf.keras.layers.Lambda(
# # lambda x: x * policy_factor, name="policy"
# )(policy)
model = tf.keras.Model(inputs=input, outputs=[policy, value])
return model
def wide_resnet_impl(input_shape, output_size):
def conv(channels,
strides,
params=dict(padding="same",
use_bias=False,
kernel_regularizer=l2(5e-4))):
def inner(x):
x = LayerNormalization()(x)
x = Activation("relu")(x)
x = Convolution2D(channels, 3, strides=strides, **params)(x)
x = Dropout(drop_rate)(x) if drop_rate > 0 else x
x = LayerNormalization()(x)
x = Activation("relu")(x)
x = Convolution2D(channels, 3, **params)(x)
return x
return inner
def resize(x, shape):
if K.int_shape(x) == shape:
return x
channels = shape[3 if K.image_data_format() ==
"channels_last" else 1]
strides = K.int_shape(x)[2] // shape[2]
return Convolution2D(channels,
1,
padding="same",
use_bias=False,
strides=strides)(x)
def block(channels, k, n, strides):
def inner(x):
for i in range(n):
x2 = conv(channels * k, strides if i == 0 else 1)(x)
x = add([resize(x, K.int_shape(x2)), x2])
return x
return inner
# According to the paper L = 6*n+4
n = int((L - 4) / 6)
group0 = Convolution2D(16,
3,
padding="same",
use_bias=False,
kernel_regularizer=l2(5e-4))
group1 = block(16, k, n, 1)
group2 = block(32, k, n, 2)
group3 = block(64, k, n, 2)
x_in = x = Input(shape=input_shape)
x = group0(x)
x = group1(x)
x = group2(x)
x = group3(x)
x = LayerNormalization()(x)
x = Activation("relu")(x)
x = GlobalAveragePooling2D()(x)
x = Dense(output_size, kernel_regularizer=l2(5e-4))(x)
y = Activation("softmax")(x)
model = Model(inputs=x_in, outputs=y)
model.compile(loss="categorical_crossentropy",
optimizer="adam",
metrics=["accuracy"])
return model
def _bn_relu(input):
"""Helper to build a BN -> relu block (by @raghakot)."""
norm = BatchNormalization(axis=CHANNEL_AXIS)(input)
return Activation("relu")(norm)