def MNet64(
input_shape=None,
depth_multiplier=1, # DepthwiseConv2D param
dropout=1e-3,
input_tensor=None,
classes=1000):
if input_tensor is None:
img_input = layers.Input(shape=input_shape)
else:
if not backend.is_keras_tensor(input_tensor):
img_input = layers.Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# Type / Stride Filter Shape Input Size
# Conv / s2 3 × 3 × 3 × 32 224 × 224 × 3 3 × 3 × 1 × 32 64 x 64 x 1 S1!
# Conv dw / s1 3 × 3 × 32 dw 112 × 112 × 32 64 x 64 x 32
# Conv / s1 1 × 1 × 32 × 64 112 × 112 × 32 64 x 64 x 32
# Conv dw / s2 3 × 3 × 64 dw 112 × 112 × 64 64 x 64 x 64 S1!
# Conv / s1 1 × 1 × 64 × 128 56 × 56 × 64 64 x 64 x 64
# Conv dw / s1 3 × 3 × 128 dw 56 × 56 × 128 64 x 64 x 128
# Conv / s1 1 × 1 × 128 × 128 56 × 56 × 128 64 x 64 x 128
# Conv dw / s2 3 × 3 × 128 dw 56 × 56 × 128 64 x 64 x 128
# Conv / s1 1 × 1 × 128 × 256 28 × 28 × 128 32 x 32 x 128
# Conv dw / s1 3 × 3 × 256 dw 28 × 28 × 256 32 x 32 x 256
# Conv / s1 1 × 1 × 256 × 256 28 × 28 × 256 32 x 32 x 256
# Conv dw / s2 3 × 3 × 256 dw 28 × 28 × 256 32 x 32 x 256
# Conv / s1 1 × 1 × 256 × 512 14 × 14 × 256 16 x 16 x 256
# 5×
# Conv dw / s1 3 × 3 × 512 dw 14 × 14 × 512 16 x 16 x 512
# Conv / s1 1 × 1 × 512 × 512 14 × 14 × 512 16 x 16 x 512
# Conv dw / s2 3 × 3 × 512 dw 14 × 14 × 512 16 x 16 x 512
# Conv / s1 1 × 1 × 512 × 1024 7 × 7 × 512 8 x 8 x 512
# Conv dw / s2 3 × 3 × 1024 dw 7 × 7 × 1024 8 x 8 x 1024
# Conv / s1 1 × 1 × 1024 × 1024 7 × 7 × 1024 8 x 8 x 1024
# Avg Pool / s1 Pool 7 × 7 7 × 7 × 1024 8 × 8 8 x 8 x 1024
# FC / s1 1024 × 1000 1 × 1 × 1024 1 × 1 × 1024
# Softmax / s1 1 × 1 × 1000 1 x 1 x NKanji
init = initializers.TruncatedNormal(stddev=0.09)
x = _conv_block(img_input, 32, strides=(1, 1),
kernel_initializer=init) # strides=(2, 2)
x = _depthwise_conv_block(x,
64,
depth_multiplier,
block_id=1,
kernel_initializer=init)
x = _depthwise_conv_block(x,
128,
depth_multiplier,
strides=(1, 1),
block_id=2,
kernel_initializer=init) # strides=(2, 2)
x = _depthwise_conv_block(x,
128,
depth_multiplier,
block_id=3,
kernel_initializer=init)
x = _depthwise_conv_block(x,
256,
depth_multiplier,
strides=(2, 2),
block_id=4,
kernel_initializer=init)
x = _depthwise_conv_block(x,
256,
depth_multiplier,
block_id=5,
kernel_initializer=init)
x = _depthwise_conv_block(x,
512,
depth_multiplier,
strides=(2, 2),
block_id=6,
kernel_initializer=init)
x = _depthwise_conv_block(x,
512,
depth_multiplier,
block_id=7,
kernel_initializer=init)
x = _depthwise_conv_block(x,
512,
depth_multiplier,
block_id=8,
kernel_initializer=init)
x = _depthwise_conv_block(x,
512,
depth_multiplier,
block_id=9,
kernel_initializer=init)
x = _depthwise_conv_block(x,
512,
depth_multiplier,
block_id=10,
kernel_initializer=init)
x = _depthwise_conv_block(x,
512,
depth_multiplier,
block_id=11,
kernel_initializer=init)
x = _depthwise_conv_block(x,
1024,
depth_multiplier,
strides=(2, 2),
block_id=12,
kernel_initializer=init)
x = _depthwise_conv_block(x,
1024,
depth_multiplier,
block_id=13,
kernel_initializer=init)
if backend.image_data_format() == 'channels_first':
shape = (1024, 1, 1)
else:
shape = (1, 1, 1024)
x = layers.GlobalAveragePooling2D()(x)
x = layers.Reshape(shape, name='reshape_1')(x)
x = layers.Dropout(dropout, name='dropout')(x)
x = layers.Conv2D(classes, (1, 1),
padding='same',
kernel_initializer=init,
name='conv_preds')(x)
x = layers.Activation('softmax', name='act_softmax')(x)
x = layers.Reshape((classes, ), name='reshape_2')(x)
model = models.Model(img_input, x, name='mobilenet')
return model
def __init__(self, shape):
self.re_rate = 0.9
self.inputs = layers.Input(shape=shape)
ac = layers.LeakyReLU(alpha=0.3)
self.f_block = layers.Conv3D(4, (5, 5, 5),
activation=ac,
kernel_regularizer=regularizers.l2(
self.re_rate),
padding='same')(self.inputs)
self.bn = layers.BatchNormalization()(self.f_block)
self.mp1 = layers.MaxPooling3D((2, 2, 2))(self.bn)
self.f_block1 = layers.Conv3D(8, (5, 5, 5),
activation=ac,
kernel_regularizer=regularizers.l2(
self.re_rate),
padding='same')(self.mp1)
self.bn = layers.BatchNormalization()(self.f_block1)
self.mp2 = layers.MaxPooling3D((2, 2, 2))(self.bn)
self.f_block2 = layers.Conv3D(16, (4, 4, 4),
activation=ac,
kernel_regularizer=regularizers.l2(
self.re_rate),
padding='same')(self.mp2)
self.f_block2 = layers.BatchNormalization()(self.f_block2)
self.mp3 = layers.MaxPooling3D((2, 2, 2))(self.f_block2)
self.f_block3 = layers.Conv3D(16, (3, 3, 3),
activation=ac,
kernel_regularizer=regularizers.l2(
self.re_rate),
padding='same')(self.mp3)
self.f_block3 = layers.BatchNormalization()(self.f_block3)
self.b_back3 = layers.Conv3D(16, (3, 3, 3),
activation=ac,
kernel_regularizer=regularizers.l2(
self.re_rate),
padding='same')(self.f_block3)
self.b_back3 = layers.BatchNormalization()(self.b_back3)
self.cat1 = layers.concatenate([self.f_block3, self.b_back3])
self.bn4 = layers.BatchNormalization()(self.cat1)
self.b_back2 = layers.Conv3D(
16, (3, 3, 3),
activation=ac,
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(layers.UpSampling3D((2, 2, 2))(self.bn4))
self.b_back2 = layers.BatchNormalization()(self.b_back2)
self.cat2 = layers.concatenate([self.mp2, self.b_back2])
self.b_back1 = layers.Conv3D(
16, (3, 3, 3),
activation=ac,
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(layers.UpSampling3D((2, 2, 2))(self.cat2))
self.b_back1 = layers.BatchNormalization()(self.b_back1)
self.cat3 = layers.concatenate([self.mp1, self.b_back1])
self.gb = layers.GlobalAveragePooling3D()(self.cat3)
# self.gb = layers.GlobalMaxPooling3D()(self.cat3)
self.dense1 = layers.Dropout(rate=0.3)(self.gb)
self.dense3 = layers.Dense(1, activation='sigmoid')(self.dense1)
self.model = keras.Model(input=self.inputs, output=self.dense3)
#Building model with best parameters
model = models.Sequential()
model.add(
layers.Conv2D(32, (5, 5), activation='relu', input_shape=(299, 299, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (5, 5), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.AveragePooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(9, activation=a))
model.add(layers.Dropout(rate=c))
model.add(layers.Dense(9, activation='softmax'))
#Compiling
model.compile(loss='categorical_crossentropy',
optimizer=e,
metrics=['accuracy'])
#Fitting
hist = model.fit(images_train,
labels_train,
batch_size=b,
epochs=d,
validation_split=0.3)
#Loss & Accuracy
def trainModel():
base_model = VGG16(weights=WEIGHTS, include_top=INCLUDE_TOP, input_shape=(HEIGHT, WIDTH, 3))
# Freeze the layers except the last 4 layers
for layer in base_model.layers[:-4]:
layer.trainable = False
# Check the trainable status of the individual layers
for layer in base_model.layers:
print(layer, layer.trainable)
# Create the model
model = models.Sequential()
# Add the vgg convolutional base model
model.add(base_model)
# Add new layers
model.add(layers.Flatten())
model.add(layers.Dense(1024, activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(3, activation='softmax'))
# Show a summary of the model. Check the number of trainable parameters
model.summary()
train_datagen = ImageDataGenerator(
rescale=1./255,
rotation_range=20,
width_shift_range=0.2,
height_shift_range=0.2,
horizontal_flip=True,
fill_mode='nearest'
)
validation_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow_from_directory(
TRAIN_DIR,
target_size=(HEIGHT, WIDTH),
batch_size=BATCH_SIZE,
class_mode='categorical')
validation_generator = validation_datagen.flow_from_directory(
VAL_DIR,
target_size=(HEIGHT, WIDTH),
batch_size=BATCH_SIZE,
class_mode='categorical',
shuffle=False)
#adam = Adam(lr=0.00001)
# Compile the model
model.compile(loss='categorical_crossentropy',
optimizer=RMSprop(lr=1e-4),
metrics=['acc'])
#model.compile(loss='categorical_crossentropy',
# optimizer=adam,
# metrics=['acc'])
filepath="./checkpoints/" + "VGG16" + "_model_weights.h5"
checkpoint = ModelCheckpoint(filepath, monitor=["acc"], verbose=1, mode='max')
callbacks_list = [checkpoint]
# Train the model
history = model.fit_generator(
train_generator,
steps_per_epoch=num_train_images // BATCH_SIZE ,
epochs=args.num_epochs,
validation_data=validation_generator,
validation_steps=num_val_images // BATCH_SIZE,
verbose=1, callbacks=callbacks_list)
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = range(len(acc))
plt.plot(epochs, acc, 'b', label='Training acc')
plt.plot(epochs, val_acc, 'r', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()
plt.figure()
plt.plot(epochs, loss, 'b', label='Training loss')
plt.plot(epochs, val_loss, 'r', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
plt.show()
imageCopy(origin_dir, train_dogs_dir, name='dog', range_num=(0, 1000))
imageCopy(origin_dir, validation_dogs_dir, name='dog', range_num=(1000, 1500))
imageCopy(origin_dir, test_dogs_dir, name='dog', range_num=(1500, 2000))
## keras model building ##
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPool2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPool2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPool2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dropout((0.5)))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
model.summary()
model.compile(optimizers.adam(lr=0.0001), loss='binary_crossentropy', metrics=['acc'])
## preprocessing ##
train_datagen = ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255) # pixel (0, 255) --> (0, 1)
# rescale / class num #
train_generator = train_datagen.flow_from_directory(train_dir,
target_size=(150, 150),
batch_size=20,
class_mode='binary')
def design_dnn(nb_features,
input_shape,
nb_levels,
conv_size,
nb_labels,
feat_mult=1,
pool_size=2,
padding='same',
activation='elu',
final_layer='dense-sigmoid',
conv_dropout=0,
conv_maxnorm=0,
nb_input_features=1,
batch_norm=False,
name=None,
prefix=None,
use_strided_convolution_maxpool=True,
nb_conv_per_level=2):
"""
"deep" cnn with dense or global max pooling layer @ end...
Could use sequential...
"""
model_name = name
if model_name is None:
model_name = 'model_1'
if prefix is None:
prefix = model_name
ndims = len(input_shape)
input_shape = tuple(input_shape)
convL = getattr(KL, 'Conv%dD' % ndims)
maxpool = KL.MaxPooling3D if len(input_shape) == 3 else KL.MaxPooling2D
if isinstance(pool_size, int):
pool_size = (pool_size, ) * ndims
# kwargs for the convolution layer
conv_kwargs = {'padding': padding, 'activation': activation}
if conv_maxnorm > 0:
conv_kwargs['kernel_constraint'] = maxnorm(conv_maxnorm)
# initialize a dictionary
enc_tensors = {}
# first layer: input
name = '%s_input' % prefix
enc_tensors[name] = KL.Input(shape=input_shape + (nb_input_features, ),
name=name)
last_tensor = enc_tensors[name]
# down arm:
# add nb_levels of conv + ReLu + conv + ReLu. Pool after each of first nb_levels - 1 layers
for level in range(nb_levels):
for conv in range(nb_conv_per_level):
if conv_dropout > 0:
name = '%s_dropout_%d_%d' % (prefix, level, conv)
enc_tensors[name] = KL.Dropout(conv_dropout)(last_tensor)
last_tensor = enc_tensors[name]
name = '%s_conv_%d_%d' % (prefix, level, conv)
nb_lvl_feats = np.round(nb_features * feat_mult**level).astype(int)
enc_tensors[name] = convL(nb_lvl_feats,
conv_size,
**conv_kwargs,
name=name)(last_tensor)
last_tensor = enc_tensors[name]
# max pool
if use_strided_convolution_maxpool:
name = '%s_strided_conv_%d' % (prefix, level)
enc_tensors[name] = convL(nb_lvl_feats,
pool_size,
**conv_kwargs,
name=name)(last_tensor)
last_tensor = enc_tensors[name]
else:
name = '%s_maxpool_%d' % (prefix, level)
enc_tensors[name] = maxpool(pool_size=pool_size,
name=name,
padding=padding)(last_tensor)
last_tensor = enc_tensors[name]
# dense layer
if final_layer == 'dense-sigmoid':
name = "%s_flatten" % prefix
enc_tensors[name] = KL.Flatten(name=name)(last_tensor)
last_tensor = enc_tensors[name]
name = '%s_dense' % prefix
enc_tensors[name] = KL.Dense(1, name=name,
activation="sigmoid")(last_tensor)
elif final_layer == 'dense-tanh':
name = "%s_flatten" % prefix
enc_tensors[name] = KL.Flatten(name=name)(last_tensor)
last_tensor = enc_tensors[name]
name = '%s_dense' % prefix
enc_tensors[name] = KL.Dense(1, name=name)(last_tensor)
last_tensor = enc_tensors[name]
# Omittting BatchNorm for now, it seems to have a cpu vs gpu problem
# https://github.com/tensorflow/tensorflow/pull/8906
# https://github.com/fchollet/keras/issues/5802
# name = '%s_%s_bn' % prefix
# enc_tensors[name] = KL.BatchNormalization(axis=batch_norm, name=name)(last_tensor)
# last_tensor = enc_tensors[name]
name = '%s_%s_tanh' % prefix
enc_tensors[name] = KL.Activation(activation="tanh",
name=name)(last_tensor)
elif final_layer == 'dense-softmax':
name = "%s_flatten" % prefix
enc_tensors[name] = KL.Flatten(name=name)(last_tensor)
last_tensor = enc_tensors[name]
name = '%s_dense' % prefix
enc_tensors[name] = KL.Dense(nb_labels,
name=name,
activation="softmax")(last_tensor)
# global max pooling layer
elif final_layer == 'myglobalmaxpooling':
name = '%s_batch_norm' % prefix
enc_tensors[name] = KL.BatchNormalization(axis=batch_norm,
name=name)(last_tensor)
last_tensor = enc_tensors[name]
name = '%s_global_max_pool' % prefix
enc_tensors[name] = KL.Lambda(_global_max_nd, name=name)(last_tensor)
last_tensor = enc_tensors[name]
name = '%s_global_max_pool_reshape' % prefix
enc_tensors[name] = KL.Reshape((1, 1), name=name)(last_tensor)
last_tensor = enc_tensors[name]
# cannot do activation in lambda layer. Could code inside, but will do extra lyaer
name = '%s_global_max_pool_sigmoid' % prefix
enc_tensors[name] = KL.Conv1D(1,
1,
name=name,
activation="sigmoid",
use_bias=True)(last_tensor)
elif final_layer == 'globalmaxpooling':
name = '%s_conv_to_featmaps' % prefix
enc_tensors[name] = KL.Conv3D(2, 1, name=name,
activation="relu")(last_tensor)
last_tensor = enc_tensors[name]
name = '%s_global_max_pool' % prefix
enc_tensors[name] = KL.GlobalMaxPooling3D(name=name)(last_tensor)
last_tensor = enc_tensors[name]
# cannot do activation in lambda layer. Could code inside, but will do extra lyaer
name = '%s_global_max_pool_softmax' % prefix
enc_tensors[name] = KL.Activation('softmax', name=name)(last_tensor)
last_tensor = enc_tensors[name]
# create the model
model = Model(inputs=[enc_tensors['%s_input' % prefix]],
outputs=[last_tensor],
name=model_name)
return model
fully_connected_size = int(input("Fully connected layer size: "))
epochs_phase_1 = int(input("Phase 1 epochs: "))
epochs_phase_2 = int(input("Phase 2 epochs: "))
batch_size = int(input("Batch size: "))
dropout_rate = float(input("Dropout rate: "))
# Clear previous layer session to match continiously constructed layer names in weights file
K.clear_session()
# Construct the classifier NN(input -> encoder -> Flatten -> FC -> output with 10 classes(0 - 9))
encoded = encoder(input_img, convolutional_layers,
convolutional_filter_size,
convolutional_filters_per_layer, dropout_rate)
flatten = layers.Flatten()(encoded)
fc = layers.Dense(fully_connected_size, activation='relu')(flatten)
dropout = layers.Dropout(rate=dropout_rate)(fc)
output_layer = layers.Dense(training_labels.num_classes(),
activation='softmax')(dropout)
classifier = Model(input_img, output_layer)
classifier.compile(loss='categorical_crossentropy',
optimizer=optimizers.Adam())
# Print it's summary
classifier.summary()
# Load encoder weights
classifier.load_weights(autoencoder_weights_file, by_name=True)
# Train phase 1: Only fully connected layer weights
def VGGNet():
image_input = Input(shape=(256, 256, 3), name='image_input')
# x = CoordinateChannel2D()(inp)
x = kl.Conv2D(filters=64,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block1_conv1')(image_input)
x = kl.Conv2D(filters=64,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block1_conv2')(x)
x = kl.Conv2D(filters=64,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block1_reduction_conv')(x)
x = kl.BatchNormalization()(x)
x = kl.Dropout(0.5)(x)
x = kl.Conv2D(filters=128,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block2_conv1')(x)
x = kl.Conv2D(filters=128,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block2_conv2')(x)
x = kl.Conv2D(filters=128,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block2_reduction_conv')(x)
x = kl.BatchNormalization()(x)
x = kl.Dropout(0.5)(x)
x = kl.Conv2D(filters=256,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block3_conv1')(x)
x = kl.Conv2D(filters=256,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block3_conv2')(x)
x = kl.Conv2D(filters=256,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block3_conv3')(x)
x = kl.Conv2D(filters=256,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block3_reduction_conv')(x)
x = kl.BatchNormalization()(x)
x = kl.Dropout(0.5)(x)
x = kl.Conv2D(filters=512,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block4_conv1')(x)
x = kl.Conv2D(filters=512,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block4_conv2')(x)
x = kl.Conv2D(filters=512,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block4_conv3')(x)
# x = CoordinateChannel2D(use_radius=True)(x)
# x, samap, g = SelfAttention(ch=512, name='self_attention')(x)
x = kl.Conv2D(filters=512,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block4_reduction_conv')(x)
x = kl.BatchNormalization()(x)
x = kl.Dropout(0.5)(x)
x = kl.Conv2D(filters=512,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block5_conv1')(x)
x = kl.Conv2D(filters=512,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block5_conv2')(x)
x = kl.Conv2D(filters=512,
kernel_size=(3, 3),
activation='relu',
padding='same',
name='block5_conv3')(x)
# x, amaps = SoftAttention(ch=512, m=32, name='soft_attention')(x)
x = kl.Conv2D(filters=512,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block5_reduction_conv')(x)
return Model(image_input, x, name='imgModel')
def DenseNet():
qw = Input(shape=(256, 256, 3), name='image_input')
qw_1 = kl.Conv2D(strides=1,
padding='valid',
activation='relu',
filters=64,
name='conv',
kernel_size=3)(qw)
qw_1 = densenet.densenet.conv_block(
x=qw_1,
growth_rate=64,
name='conv_1',
)
qw_2 = densenet.densenet.dense_block(qw_1, blocks=1, name='block_1')
qw_2 = kl.BatchNormalization()(qw_2)
qw_2 = kl.Activation('relu')(qw_2)
qw_2 = kl.Conv2D(filters=64,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block1_reduction_conv')(qw_2)
qw_2 = kl.Dropout(0.5)(qw_2)
qw_2 = densenet.densenet.dense_block(qw_2, blocks=1, name='block_2')
qw_2 = kl.BatchNormalization()(qw_2)
qw_2 = kl.Activation('relu')(qw_2)
# qw_2 = kl.MaxPool2D(pool_size=2)(qw_2)
qw_2 = kl.Conv2D(filters=128,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block2_reduction_conv')(qw_2)
qw_2 = kl.Dropout(0.5)(qw_2)
qw_2 = densenet.densenet.dense_block(qw_2, blocks=1, name='block_3')
qw_2 = kl.BatchNormalization()(qw_2)
qw_2 = kl.Activation('relu')(qw_2)
# qw_2 = kl.MaxPool2D(pool_size=2)(qw_2)
qw_2 = kl.Conv2D(filters=256,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block3_reduction_conv')(qw_2)
qw_2 = kl.Dropout(0.5)(qw_2)
qw_2 = densenet.densenet.dense_block(qw_2, blocks=1, name='block_4')
qw_2 = kl.BatchNormalization()(qw_2)
qw_2 = kl.Activation('relu')(qw_2)
# qw_2 = kl.MaxPool2D(pool_size=2)(qw_2)
qw_2 = kl.Conv2D(filters=512,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block4_reduction_conv')(qw_2)
qw_2 = kl.Dropout(0.5)(qw_2)
qw_2 = densenet.densenet.dense_block(qw_2, blocks=1, name='block_5')
qw_2 = kl.BatchNormalization()(qw_2)
qw_2 = kl.Activation('relu')(qw_2)
# qw_2 = kl.MaxPool2D(pool_size=2)(qw_2)
qw_2 = kl.Conv2D(filters=1024,
kernel_size=2,
strides=2,
activation='relu',
padding='same',
name='block5_reduction_conv')(qw_2)
# qw_2 = kl.Dropout(0.5)(qw_2)
qw_2 = kl.BatchNormalization()(qw_2)
qw_2 = kl.Activation('relu')(qw_2)
return Model(qw, qw_2, name='imgModel')
def get_model(outputs, gpus=1):
model = None
if os.path.isfile(WEIGHTS_FILE) and False:
model = models.load_model(WEIGHTS_FILE)
print('model loaded')
else:
inp = layers.Input(shape=(
8,
8,
12,
))
flat_inp = layers.Flatten()(inp)
y = inp
y = residual_block(y, 128, _project_shortcut=True)
for i in range(25):
y = residual_block(y, 128, _project_shortcut=False)
y = layers.Flatten()(y)
inp2 = layers.Input(shape=(4, ))
y = layers.concatenate([y, flat_inp, inp2])
y = layers.Dense(1024,
activation='relu',
kernel_regularizer=regularizers.l2(WEIGHT_DECAY))(y)
y = layers.Dropout(0.1)(y)
y_policy = layers.Dropout(0.2)(y)
y_policy = layers.Dense(
1024,
activation='relu',
kernel_regularizer=regularizers.l2(WEIGHT_DECAY))(y_policy)
y_policy = layers.Dropout(0.1)(y_policy)
y_policy = layers.Dense(
2048,
activation='relu',
kernel_regularizer=regularizers.l2(WEIGHT_DECAY))(y_policy)
y_policy = layers.Dropout(0.1)(y_policy)
y_policy = layers.Dense(
outputs,
activation='relu',
kernel_regularizer=regularizers.l2(WEIGHT_DECAY))(y_policy)
y_policy = layers.Activation("softmax")(y_policy)
y_value = layers.Dropout(0.2)(y)
y_value = layers.Dense(
1024,
activation='relu',
kernel_regularizer=regularizers.l2(WEIGHT_DECAY))(y_value)
y_value = layers.Dropout(0.1)(y_value)
y_value = layers.Dense(
2048,
activation='relu',
kernel_regularizer=regularizers.l2(WEIGHT_DECAY))(y_value)
y_value = layers.Dropout(0.1)(y_value)
y_value = layers.Dense(
2,
activation='relu',
kernel_regularizer=regularizers.l2(WEIGHT_DECAY))(y_value)
y_value = layers.Activation("softmax")(y_value)
model = models.Model(inputs=[inp, inp2], outputs=[y_policy, y_value])
# model.summary()
model.compile(optimizer=optimizers.SGD(lr=0.001, momentum=0.6),
loss='categorical_crossentropy',
loss_weights=[1., 1.],
metrics=['accuracy'])
if os.path.isfile(WEIGHTS_FILE):
model.load_weights(WEIGHTS_FILE)
return model, model
textNet = textNet()
textNet.summary()
# In[301]:
SVG(
model_to_dot(textNet, show_layer_names=True,
show_shapes=True).create(prog='dot', format='svg'))
# # Define Complete Joint Network
# In[302]:
image_input = Input(shape=(256, 256, 3), name='image_input')
image_features = imgNet(image_input)
image_features = kl.Dropout(0.1)(image_features)
image_features = kl.Flatten()(image_features)
image_features = kl.Dense(512, activation='relu')(image_features)
words_input = kl.Input(shape=(max_wlen, ), name='words_input')
text_features = textNet(words_input)
text_features = kl.Dropout(0.1)(text_features)
text_features = kl.Dense(512, activation='relu')(text_features)
g = kl.Concatenate()([image_features, text_features])
g = kl.Dense(512, activation='relu')(g)
target = Dense(vocab_size + 1, activation='softmax', name='target_word')(g)
model = Model([image_input, words_input], target)
# In[303]:
def VGG16():
# Determine proper input shape
input_shape = (128, 128, 200)
img_input = layers.Input(shape=input_shape)
# Block 1
x = layers.Conv2D(64, (3, 3),
activation='relu',
padding='same',
name='block1_conv1')(img_input)
x = layers.Conv2D(64, (3, 3),
activation='relu',
padding='same',
name='block1_conv2')(x)
x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(x)
# Block 2
x = layers.Conv2D(128, (3, 3),
activation='relu',
padding='same',
name='block2_conv1')(x)
x = layers.Conv2D(128, (3, 3),
activation='relu',
padding='same',
name='block2_conv2')(x)
x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(x)
# Block 3
x = layers.Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv1')(x)
x = layers.Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv2')(x)
x = layers.Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv3')(x)
x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool')(x)
# Block 4
x = layers.Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv1')(x)
x = layers.Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv2')(x)
x = layers.Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv3')(x)
x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool')(x)
# Block 5
x = layers.Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv1')(x)
x = layers.Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv2')(x)
x = layers.Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv3')(x)
x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block5_pool')(x)
# Classification block
x = layers.Flatten(name='flatten')(x)
x = layers.Dropout(0.3)(x)
x = layers.Dense(4096, activation='relu', name='fc1')(x)
x = layers.Dropout(0.3)(x)
x = layers.Dense(4096, activation='relu', name='fc2')(x)
#x = layers.Dense(2, name='predictions')(x)
inputs = img_input
# Create model.
#model = models.Model(inputs, x, name='vgg16')
y1 = layers.Dense(11, activation='softmax')(x)
y2 = layers.Dense(11, activation='softmax')(x)
model = Model(inputs=inputs, outputs=[y1, y2])
model.load_weights("weight/model_2_v1_tf.hdf5")
adam = Adam(lr=0.0001)
model.compile(loss='categorical_crossentropy',
optimizer=adam,
metrics=['accuracy'])
return model
trains_l = trainig_data_labels[0:trains_nums]
print(trains_d.shape)
print(trains_l.shape)
val_d = trainig_data[trains_nums:val_nums + trains_nums]
val_l = trainig_data_labels[trains_nums:val_nums + trains_nums]
# graph
input_tensor = Input(shape=(13, ))
x = layers.Dense(
64, activation='relu',
kernel_regularizer=keras.regularizers.l2(0.001))(input_tensor)
x = layers.Dense(128,
activation='relu',
kernel_regularizer=keras.regularizers.l2(0.001))(x)
x = layers.Dropout(0.5)(x)
x = layers.Dense(64,
activation='relu',
kernel_regularizer=keras.regularizers.l2(0.001))(x)
output_tensor = layers.Dense(14, activation='softmax')(x)
model = Model(input_tensor, output_tensor)
model.summary()
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
callback = [
keras.callbacks.TensorBoard(
log_dir="log_dir",
from keras import layers, models
from loader import prepare_uic_data
X, Y = prepare_uic_data()
X = X.reshape(*X.shape, 1)
model = models.Sequential()
model.add(layers.Conv2D(32, (5, 5), activation='relu',
input_shape=X.shape[1:]))
model.add(layers.MaxPooling2D())
model.add(layers.Conv2D(32, (5, 5), activation='relu'))
model.add(layers.MaxPooling2D())
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dropout(.2))
model.add(layers.Dense(2, activation='softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='sgd',
metrics=['accuracy'])
model.fit(X, Y, validation_split=.2, epochs=10)
model.save('uic_cnn.h5')
def conv_enc(nb_features,
input_shape,
nb_levels,
conv_size,
name=None,
prefix=None,
feat_mult=1,
pool_size=2,
dilation_rate_mult=1,
padding='same',
activation='elu',
layer_nb_feats=None,
use_residuals=False,
nb_conv_per_level=2,
conv_dropout=0,
batch_norm=None):
"""
Fully Convolutional Encoder
"""
# naming
model_name = name
if prefix is None:
prefix = model_name
# volume size data
ndims = len(input_shape) - 1
input_shape = tuple(input_shape)
if isinstance(pool_size, int):
pool_size = (pool_size, ) * ndims
# prepare layers
convL = getattr(KL, 'Conv%dD' % ndims)
conv_kwargs = {'padding': padding, 'activation': activation}
maxpool = getattr(KL, 'MaxPooling%dD' % ndims)
# first layer: input
name = '%s_input' % prefix
last_tensor = KL.Input(shape=input_shape, name=name)
input_tensor = last_tensor
# down arm:
# add nb_levels of conv + ReLu + conv + ReLu. Pool after each of first nb_levels - 1 layers
lfidx = 0
for level in range(nb_levels):
lvl_first_tensor = last_tensor
nb_lvl_feats = np.round(nb_features * feat_mult**level).astype(int)
conv_kwargs['dilation_rate'] = dilation_rate_mult**level
for conv in range(nb_conv_per_level):
if layer_nb_feats is not None:
nb_lvl_feats = layer_nb_feats[lfidx]
lfidx += 1
name = '%s_conv_downarm_%d_%d' % (prefix, level, conv)
if conv < (nb_conv_per_level - 1) or (not use_residuals):
last_tensor = convL(nb_lvl_feats,
conv_size,
**conv_kwargs,
name=name)(last_tensor)
else: # no activation
last_tensor = convL(nb_lvl_feats,
conv_size,
padding=padding,
name=name)(last_tensor)
if conv_dropout > 0:
# conv dropout along feature space only
name = '%s_dropout_downarm_%d_%d' % (prefix, level, conv)
noise_shape = [None, *[1] * ndims, nb_lvl_feats]
last_tensor = KL.Dropout(conv_dropout,
noise_shape=noise_shape)(last_tensor)
if use_residuals:
convarm_layer = last_tensor
# the "add" layer is the original input
# However, it may not have the right number of features to be added
nb_feats_in = lvl_first_tensor.get_shape()[-1]
nb_feats_out = convarm_layer.get_shape()[-1]
add_layer = lvl_first_tensor
if nb_feats_in > 1 and nb_feats_out > 1 and (nb_feats_in !=
nb_feats_out):
name = '%s_expand_down_merge_%d' % (prefix, level)
last_tensor = convL(nb_lvl_feats,
conv_size,
**conv_kwargs,
name=name)(lvl_first_tensor)
add_layer = last_tensor
if conv_dropout > 0:
name = '%s_dropout_down_merge_%d_%d' % (prefix, level,
conv)
noise_shape = [None, *[1] * ndims, nb_lvl_feats]
last_tensor = KL.Dropout(
conv_dropout, noise_shape=noise_shape)(last_tensor)
name = '%s_res_down_merge_%d' % (prefix, level)
last_tensor = KL.add([add_layer, convarm_layer], name=name)
name = '%s_res_down_merge_act_%d' % (prefix, level)
last_tensor = KL.Activation(activation, name=name)(last_tensor)
if batch_norm is not None:
name = '%s_bn_down_%d' % (prefix, level)
last_tensor = KL.BatchNormalization(axis=batch_norm,
name=name)(last_tensor)
# max pool if we're not at the last level
if level < (nb_levels - 1):
name = '%s_maxpool_%d' % (prefix, level)
last_tensor = maxpool(pool_size=pool_size,
name=name,
padding=padding)(last_tensor)
# create the model and return
model = Model(inputs=input_tensor, outputs=[last_tensor], name=model_name)
return model
# add START and END to lines
lines = [['START'] + l + ['END'] for l in lines]
# force all lines to be same length
maxlen = 0
for l in lines:
if len(l) > maxlen:
maxlen = len(l)
for i in range(len(lines)):
if len(lines[i]) < maxlen:
lines[i] += ['BLANK'] * (maxlen - len(lines[i]))
# condense list of paragraphs into an np tensor
# dimensions: examples/sentences, character vectors, characters 1/0s
data = np.zeros((len(lines), maxlen, num_chars))
for i, line in enumerate(lines):
for j, c in enumerate(line):
data[i][j][char_to_ind[c]] = 1
# split data into inputs and outputs
seq_len = 100
X = np.zeros((0, seq_len, num_chars))
# create LSTM model
lstm_input = kl.Input(shape=[maxlen, num_chars])
H = kl.LSTM(256)(lstm_input)
H = kl.Dropout(0.2)(H)
lstm_output = kl.Dense(num_chars, activation='softmax')(H)
lstm = km.Model(lstm_input, lstm_output)
lstm.compile(loss="categorical_crossentropy", optimizer="adam")
def conv_dec(nb_features,
input_shape,
nb_levels,
conv_size,
nb_labels,
name=None,
prefix=None,
feat_mult=1,
pool_size=2,
use_skip_connections=False,
padding='same',
dilation_rate_mult=1,
activation='elu',
use_residuals=False,
final_pred_activation='softmax',
nb_conv_per_level=2,
layer_nb_feats=None,
batch_norm=None,
conv_dropout=0,
input_model=None):
"""
Fully Convolutional Decoder
Parameters:
...
use_skip_connections (bool): if true, turns an Enc-Dec to a U-Net.
If true, input_tensor and tensors are required.
It assumes a particular naming of layers. conv_enc...
"""
# naming
model_name = name
if prefix is None:
prefix = model_name
# if using skip connections, make sure need to use them.
if use_skip_connections:
assert input_model is not None, "is using skip connections, tensors dictionary is required"
# first layer: input
input_name = '%s_input' % prefix
if input_model is None:
input_tensor = KL.Input(shape=input_shape, name=input_name)
last_tensor = input_tensor
else:
input_tensor = input_model.input
last_tensor = input_model.output
input_shape = last_tensor.shape.as_list()[1:]
# vol size info
ndims = len(input_shape) - 1
input_shape = tuple(input_shape)
if isinstance(pool_size, int):
if ndims > 1:
pool_size = (pool_size, ) * ndims
# prepare layers
convL = getattr(KL, 'Conv%dD' % ndims)
conv_kwargs = {'padding': padding, 'activation': activation}
upsample = getattr(KL, 'UpSampling%dD' % ndims)
# up arm:
# nb_levels - 1 layers of Deconvolution3D
# (approx via up + conv + ReLu) + merge + conv + ReLu + conv + ReLu
lfidx = 0
for level in range(nb_levels - 1):
nb_lvl_feats = np.round(nb_features *
feat_mult**(nb_levels - 2 - level)).astype(int)
conv_kwargs['dilation_rate'] = dilation_rate_mult**(nb_levels - 2 -
level)
# upsample matching the max pooling layers size
name = '%s_up_%d' % (prefix, nb_levels + level)
last_tensor = upsample(size=pool_size, name=name)(last_tensor)
up_tensor = last_tensor
# merge layers combining previous layer
# TODO: add Cropping3D or Cropping2D if 'valid' padding
if use_skip_connections:
conv_name = '%s_conv_downarm_%d_%d' % (
prefix, nb_levels - 2 - level, nb_conv_per_level - 1)
cat_tensor = input_model.get_layer(conv_name).output
name = '%s_merge_%d' % (prefix, nb_levels + level)
last_tensor = KL.concatenate([cat_tensor, last_tensor],
axis=ndims + 1,
name=name)
# convolution layers
for conv in range(nb_conv_per_level):
if layer_nb_feats is not None:
nb_lvl_feats = layer_nb_feats[lfidx]
lfidx += 1
name = '%s_conv_uparm_%d_%d' % (prefix, nb_levels + level, conv)
if conv < (nb_conv_per_level - 1) or (not use_residuals):
last_tensor = convL(nb_lvl_feats,
conv_size,
**conv_kwargs,
name=name)(last_tensor)
else:
last_tensor = convL(nb_lvl_feats,
conv_size,
padding=padding,
name=name)(last_tensor)
if conv_dropout > 0:
name = '%s_dropout_uparm_%d_%d' % (prefix, level, conv)
noise_shape = [None, *[1] * ndims, nb_lvl_feats]
last_tensor = KL.Dropout(conv_dropout,
noise_shape=noise_shape)(last_tensor)
# residual block
if use_residuals:
# the "add" layer is the original input
# However, it may not have the right number of features to be added
add_layer = up_tensor
nb_feats_in = add_layer.get_shape()[-1]
nb_feats_out = last_tensor.get_shape()[-1]
if nb_feats_in > 1 and nb_feats_out > 1 and (nb_feats_in !=
nb_feats_out):
name = '%s_expand_up_merge_%d' % (prefix, level)
add_layer = convL(nb_lvl_feats,
conv_size,
**conv_kwargs,
name=name)(add_layer)
if conv_dropout > 0:
name = '%s_dropout_up_merge_%d_%d' % (prefix, level, conv)
noise_shape = [None, *[1] * ndims, nb_lvl_feats]
last_tensor = KL.Dropout(
conv_dropout, noise_shape=noise_shape)(last_tensor)
name = '%s_res_up_merge_%d' % (prefix, level)
last_tensor = KL.add([last_tensor, add_layer], name=name)
name = '%s_res_up_merge_act_%d' % (prefix, level)
last_tensor = KL.Activation(activation, name=name)(last_tensor)
if batch_norm is not None:
name = '%s_bn_up_%d' % (prefix, level)
last_tensor = KL.BatchNormalization(axis=batch_norm,
name=name)(last_tensor)
# Compute likelyhood prediction (no activation yet)
name = '%s_likelihood' % prefix
last_tensor = convL(nb_labels, 1, activation=None, name=name)(last_tensor)
like_tensor = last_tensor
# output prediction layer
# we use a softmax to compute P(L_x|I) where x is each location
if final_pred_activation == 'softmax':
print("using final_pred_activation %s for %s" %
(final_pred_activation, model_name))
name = '%s_prediction' % prefix
softmax_lambda_fcn = lambda x: keras.activations.softmax(
x, axis=ndims + 1)
pred_tensor = KL.Lambda(softmax_lambda_fcn, name=name)(last_tensor)
# otherwise create a layer that does nothing.
else:
name = '%s_prediction' % prefix
pred_tensor = KL.Activation('linear', name=name)(like_tensor)
# create the model and retun
model = Model(inputs=input_tensor, outputs=pred_tensor, name=model_name)
return model
def add_common_layers(y):
y = layers.advanced_activations.LeakyReLU()(y)
y = layers.Dropout(0.25)(y)
return y
def model(reader):
assert reader.channeltypes == [ChannelType.CHAR, ChannelType.WORD]
text_1_char = L.Input(shape=(None, ), dtype='int32')
text_1_word = L.Input(shape=(None, ), dtype='int32')
text_2_char = L.Input(shape=(None, ), dtype='int32')
text_2_word = L.Input(shape=(None, ), dtype='int32')
word_mapping = reader.channels[1].vocabulary_map
word_weights = gew.generate_embedding_weights(word_mapping)
word_number = word_weights.shape[0]
word_embedding_size = word_weights.shape[1]
char_embedding = L.Embedding(
len(reader.channels[0].vocabulary_above_cutoff) + 2, 5)
word_embedding = L.Embedding(output_dim=word_embedding_size,
input_dim=word_number,
trainable=False,
weights=[word_weights])
# Embed all inputs.
text_1_char_emb = char_embedding(text_1_char)
text_1_word_emb = word_embedding(text_1_word)
text_2_char_emb = char_embedding(text_2_char)
text_2_word_emb = word_embedding(text_2_word)
# Define convolutions.
char_conv8 = L.Convolution1D(filters=200,
kernel_size=8,
strides=1,
activation='relu')
char_conv4 = L.Convolution1D(filters=200,
kernel_size=4,
strides=1,
activation='relu')
word_conv8 = L.Convolution1D(filters=100,
kernel_size=8,
strides=1,
activation='relu')
# Convolve all inputs and pool them to a representation of the texts.
text_1_char_repr_8 = L.GlobalMaxPooling1D()(char_conv8(text_1_char_emb))
text_1_char_repr_4 = L.GlobalMaxPooling1D()(char_conv4(text_1_char_emb))
text_1_word_repr_8 = L.GlobalMaxPooling1D()(word_conv8(text_1_word_emb))
text_2_char_repr_8 = L.GlobalMaxPooling1D()(char_conv8(text_2_char_emb))
text_2_char_repr_4 = L.GlobalMaxPooling1D()(char_conv4(text_2_char_emb))
text_2_word_repr_8 = L.GlobalMaxPooling1D()(word_conv8(text_2_word_emb))
# Representation of texts are concatenation of all representations.
text_1_repr = L.Concatenate()(
[text_1_char_repr_8, text_1_char_repr_4, text_1_word_repr_8])
text_2_repr = L.Concatenate()(
[text_2_char_repr_8, text_2_char_repr_4, text_2_word_repr_8])
# Find the difference between the texts and learn on it.
abs_diff = L.merge(inputs=[text_1_repr, text_2_repr],
mode=lambda x: abs(x[0] - x[1]),
output_shape=lambda x: x[0],
name='absolute_difference')
dense1 = L.Dense(500, activation='relu')(abs_diff)
dense2 = L.Dense(500, activation='relu')(dense1)
pruned = L.Dropout(0.3)(dense2)
output = L.Dense(2, activation='softmax', name='output')(pruned)
model = Model(inputs=[text_1_char, text_1_word, text_2_char, text_2_word],
outputs=output)
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
return model
def Conv2DMultiTaskIn1(x_train, y_train, ddg_train, x_test, y_test, ddg_test,
x_val, y_val, ddg_val, class_weights_dict, modeldir):
K.clear_session()
summary = False
verbose = 0
# setHyperParams------------------------------------------------------------------------------------------------
batch_size = 64
epochs = 200
lr = 0.0049
optimizer = 'sgd'
activator = 'elu'
basic_conv2D_layers = 1
basic_conv2D_filter_num = 16
loop_dilation2D_layers = 2
loop_dilation2D_filter_num = 64
loop_dilation2D_dropout_rate = 0.2008
dilation_lower = 2
dilation_upper = 16
reduce_layers = 3 # conv 3 times: 120 => 60 => 30 => 15
# print(reduce_layers)
reduce_conv2D_filter_num = 16
reduce_conv2D_dropout_rate = 0.1783
residual_stride = 2
dense1_num = 128
dense2_num = 32
drop_num = 0.2605
kernel_size = (3, 3)
pool_size = (2, 2)
initializer = 'random_uniform'
padding_style = 'same'
loss_type = 'mse'
metrics = ('mae', pearson_r)
my_callbacks = [
callbacks.ReduceLROnPlateau(
monitor='val_loss',
factor=0.5,
patience=5,
)
]
if lr > 0:
if optimizer == 'adam':
chosed_optimizer = optimizers.Adam(lr=lr)
elif optimizer == 'sgd':
chosed_optimizer = optimizers.SGD(lr=lr)
elif optimizer == 'rmsprop':
chosed_optimizer = optimizers.RMSprop(lr=lr)
# build --------------------------------------------------------------------------------------------------------
## basic Conv2D
input_layer = Input(shape=x_train.shape[1:])
y = layers.Conv2D(basic_conv2D_filter_num,
kernel_size,
padding=padding_style,
kernel_initializer=initializer,
activation=activator)(input_layer)
y = layers.BatchNormalization(axis=-1)(y)
if basic_conv2D_layers == 2:
y = layers.Conv2D(basic_conv2D_filter_num,
kernel_size,
padding=padding_style,
kernel_initializer=initializer,
activation=activator)(y)
y = layers.BatchNormalization(axis=-1)(y)
## loop with Conv2D with dilation (padding='same')
for _ in range(loop_dilation2D_layers):
y = layers.Conv2D(loop_dilation2D_filter_num,
kernel_size,
padding=padding_style,
dilation_rate=dilation_lower,
kernel_initializer=initializer,
activation=activator)(y)
y = layers.BatchNormalization(axis=-1)(y)
y = layers.Dropout(loop_dilation2D_dropout_rate)(y)
dilation_lower *= 2
if dilation_lower > dilation_upper:
dilation_lower = 2
## Conv2D with dilation (padding='valaid') and residual block to reduce dimention.
for _ in range(reduce_layers):
y = layers.Conv2D(reduce_conv2D_filter_num,
kernel_size,
padding=padding_style,
kernel_initializer=initializer,
activation=activator)(y)
y = layers.BatchNormalization(axis=-1)(y)
y = layers.Dropout(reduce_conv2D_dropout_rate)(y)
y = layers.MaxPooling2D(pool_size, padding=padding_style)(y)
residual = layers.Conv2D(reduce_conv2D_filter_num,
1,
strides=residual_stride,
padding='same')(input_layer)
y = layers.add([y, residual])
residual_stride *= 2
## flat & dense
y = layers.Flatten()(y)
y = layers.Dense(dense1_num, activation=activator)(y)
y = layers.BatchNormalization(axis=-1)(y)
y = layers.Dropout(drop_num)(y)
y = layers.Dense(dense2_num, activation=activator)(y)
y = layers.BatchNormalization(axis=-1)(y)
y = layers.Dropout(drop_num)(y)
output_layer = layers.Dense(1)(y)
model = models.Model(inputs=input_layer, outputs=output_layer)
if summary:
model.summary()
model.compile(
optimizer=chosed_optimizer,
loss=loss_type,
metrics=list(metrics) # accuracy
)
K.set_session(tf.Session(graph=model.output.graph))
init = K.tf.global_variables_initializer()
K.get_session().run(init)
result = model.fit(
x=x_train,
y=ddg_train,
batch_size=batch_size,
epochs=epochs,
verbose=verbose,
callbacks=my_callbacks,
validation_data=(x_test, ddg_test),
shuffle=True,
)
# print('\n----------History:\n%s'%result.history)
model_json = model.to_json()
with open(modeldir + '/model.json', 'w') as json_file:
json_file.write(model_json)
return model, result.history
# 将28*28的方阵变成向量
train_images = train_images.reshape((60000, 28 * 28)).astype('float')
test_images = test_images.reshape((10000, 28 * 28)).astype('float')
# 1 -> [0 1 0 0 0...]
train_labels = to_categorical(train_labels)
test_labels = to_categorical(test_labels)
network = models.Sequential()
network.add(
layers.Dense(
units=128,
activation='relu',
input_shape=(28 * 28, ),
kernel_regularizer=regularizers.l1(0.0001))) # 正则化 lambda=0.0001
network.add(layers.Dropout(0.01)) # 防止过拟合
network.add(
layers.Dense(units=32,
activation='relu',
kernel_regularizer=regularizers.l1(0.0001)))
network.add(layers.Dropout(0.01))
network.add(layers.Dense(units=10, activation='softmax'))
# 优化器：RMSprop 损失函数：交叉熵损失函数
network.compile(optimizer=RMSprop(lr=0.001),
loss='categorical_crossentropy',
metrics=['accuracy'])
# 训练网络, epochs训练回合， batch_size每次训练给的数据
network.fit(train_images, train_labels, epochs=20, batch_size=128, verbose=2)
input_shape=(h, w, 1 if bw == 0 else 3),
padding='valid'))
model.add(layers.BatchNormalization())
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (5, 5), activation='relu'))
model.add(layers.BatchNormalization())
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.BatchNormalization())
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(256, (3, 3), activation='relu'))
model.add(layers.BatchNormalization())
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(2048, activation='relu'))
model.add(layers.Dropout(0.2))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dropout(0.2))
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dropout(0.2))
model.add(layers.Dense(1, activation='sigmoid'))
model.compile(optimizer='rmsprop',
loss='binary_crossentropy',
metrics=['acc'])
print(model.summary())
history = model.fit(train_set,
train_label,
validation_data=(val_set, val_label),
epochs=int(epocs),
X_train, X_test, y_train, y_test = train_test_split(headlines,
targets,
test_size=0.2,
random_state=0)
return X_train, X_test, y_train, y_test
if __name__ == '__main__':
X_train, X_test, y_train, y_test = preprocess()
clf = models.Sequential()
clf.add(layers.Dense(256, activation='relu', input_shape=(X_train.shape[1],)))
clf.add(layers.Dense(256, activation='relu'))
clf.add(layers.Dense(256, activation='relu'))
clf.add(layers.Dropout(rate=0.2))
clf.add(layers.Dense(256, activation='relu'))
clf.add(layers.Dropout(rate=0.2))
clf.add(layers.Dense(41)) # 41是文本类型数
clf.add(layers.Activation('softmax'))
clf.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])
clf.fit(X_train, y_train, epochs=35, batch_size=100, verbose=2)
# test model
aa = clf.predict(X_test) # 具体看测试集预测情况
for i in range(300):
print(aa[i].argmax(), y_test[i].argmax(), '\n')
print("RESULT ON TEST: ", clf.evaluate(X_test, y_test, batch_size=1000))
def multi_net(img_rows, img_cols, color_type, num_classes=None):
small = L.Input(shape=(img_rows, img_cols, color_type), name='small')
x = conv_block(small, 5)
x = L.Conv2D(36, (5, 5), activation='relu')(small)
x = L.Conv2D(48, (3, 3), activation='relu')(x)
x = L.Flatten()(x)
x1 = L.Dense(256)(x)
x = L.Dense(512, activation='relu')(x)
x = L.Dropout(0.4)(x)
x = L.Dense(512, activation='relu')(x)
x = L.Dropout(0.4)(x)
logits1 = L.Dense(num_classes, activation='softmax')(x)
medium = L.Input(shape=(img_rows * 2, img_cols * 2, color_type),
name='medium')
x = conv_block(medium, 5)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = conv_block(x, 3)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Flatten()(x)
x2 = L.Dense(256)(x)
x = L.Dense(512, activation='relu')(x)
x = L.Dropout(0.4)(x)
x = L.Dense(512, activation='relu')(x)
x = L.Dropout(0.4)(x)
logits2 = L.Dense(num_classes, activation='softmax')(x)
large = L.Input(shape=(img_rows * 3, img_cols * 3, color_type),
name='large')
x = conv_block(large, 5)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = conv_block(x, 3)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Conv2D(16, (3, 3), activation='relu')(x)
x = L.Flatten()(x)
x3 = L.Dense(256)(x)
x = L.Dense(512, activation='relu')(x)
x = L.Dropout(0.4)(x)
x = L.Dense(512, activation='relu')(x)
x = L.Dropout(0.4)(x)
logits3 = L.Dense(num_classes, activation='softmax')(x)
# # combine all patch
# merge0=L.concatenate([x1, x2, x3],axis=-1)
# merge1=L.Dense(1024)(merge0)
# merge2=L.Activation('relu')(merge1)
# merge2 = L.Dropout(0.4)(merge2)
# merge3=L.Dense(512)(merge2)
# merge3=L.Activation('relu')(merge3)
# merge3 = L.Dropout(0.4)(merge3)
# logits=L.Dense(num_classes,activation='softmax')(merge0)
logits = L.average([logits1, logits2, logits3])
new_model = K.models.Model([small, medium, large], logits)
sgd = K.optimizers.SGD(lr=0.001, momentum=0.99, decay=1e-4)
new_model.compile(optimizer=sgd,
loss='categorical_crossentropy',
metrics=['acc'])
return new_model
def __init__(self, shape):
self.re_rate = 0.9
self.inputs = layers.Input(shape=shape)
self.f_block = layers.Conv3D(4, (3, 3, 3), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.inputs)
self.bn = layers.BatchNormalization()(self.f_block)
self.mp1 = layers.MaxPooling3D((2, 2, 2))(self.bn)
self.f_block1 = layers.Conv3D(16, (3, 3, 3), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.mp1)
self.bn = layers.BatchNormalization()(self.f_block1)
self.f_block1 = layers.Conv3D(16, (1, 1, 1), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.bn)
self.f_block1 = layers.BatchNormalization()(self.f_block1)
self.mp2 = layers.MaxPooling3D((2, 2, 2))(self.f_block1)
self.f_block2 = layers.Conv3D(32, (3, 3, 3), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.mp2)
self.f_block2 = layers.BatchNormalization()(self.f_block2)
self.f_block2 = layers.Conv3D(32, (1, 1, 1), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.f_block2)
self.f_block2 = layers.BatchNormalization()(self.f_block2)
self.mp3 = layers.MaxPooling3D((2, 2, 2))(self.f_block2)
self.f_block3 = layers.Conv3D(64, (3, 3, 3), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.mp3)
self.f_block3 = layers.BatchNormalization()(self.f_block3)
self.f_block3 = layers.Conv3D(64, (1, 1, 1), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.f_block3)
self.f_block3 = layers.BatchNormalization()(self.f_block3)
# self.mp4 = layers.MaxPooling3D((2, 2, 2))(self.f_block3)
self.b_back3 = layers.Conv3D(64, (3, 3, 3), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.f_block3)
self.b_back3 = layers.BatchNormalization()(self.b_back3)
self.b_back3 = layers.Conv3D(64, (1, 1, 1), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.b_back3)
self.b_back3 = layers.BatchNormalization()(self.b_back3)
self.cat1 = layers.concatenate([self.f_block3, self.b_back3])
self.bn4 = layers.BatchNormalization()(self.cat1)
self.b_back2 = layers.Conv3D(64, (3, 3, 3), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(layers.UpSampling3D((2, 2, 2))(self.bn4))
self.b_back2 = layers.BatchNormalization()(self.b_back2)
self.b_back2 = layers.Conv3D(64, (1, 1, 1), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.b_back2)
self.b_back2 = layers.BatchNormalization()(self.b_back2)
self.cat2 = layers.concatenate([self.mp2, self.b_back2])
self.bn = layers.BatchNormalization()(self.cat2)
self.b_back1 = layers.Conv3D(32, (3, 3, 3), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(layers.UpSampling3D((2, 2, 2))(self.bn))
self.b_back1 = layers.BatchNormalization()(self.b_back1)
self.b_back1 = layers.Conv3D(32, (1, 1, 1), activation='relu',
kernel_regularizer=regularizers.l2(self.re_rate),
padding='same')(self.b_back1)
self.b_back1 = layers.BatchNormalization()(self.b_back1)
self.cat3 = layers.concatenate([self.mp1, self.b_back1])
self.gb = layers.GlobalAveragePooling3D()(self.cat3)
self.drop = layers.Dropout(rate=0.7)(self.gb)
self.dense1 = layers.Dense(128, activation='relu')(self.drop)
self.dense1 = layers.Dropout(rate=0.6)(self.dense1)
self.dense2 = layers.Dense(64, activation='relu')(self.dense1)
self.dense3 = layers.Dense(4, activation='softmax')(self.dense2)
self.model = keras.Model(input=self.inputs, output=self.dense3)
def __init__(self, name=None):
super().__init__(name=name)
self.d1 = layers.Dense(1000)
self.d2 = layers.Dense(1000)
self.dropout = layers.Dropout(0.1)
# In[2]:
discriminator_input = layers.Input(shape=(height, width, channels))
x = layers.Conv2D(128, 3)(discriminator_input)
x = layers.LeakyReLU()(x)
x = layers.Conv2D(128, 4, strides=2)(x)
x = layers.LeakyReLU()(x)
x = layers.Conv2D(128, 4, strides=2)(x)
x = layers.LeakyReLU()(x)
x = layers.Conv2D(128, 4, strides=2)(x)
x = layers.LeakyReLU()(x)
x = layers.Flatten()(x)
# One dropout layer - important trick!
x = layers.Dropout(0.4)(x)
# Classification layer
x = layers.Dense(1, activation='sigmoid')(x)
discriminator = keras.models.Model(discriminator_input, x)
discriminator.summary()
# To stabilize training, we use learning rate decay
# and gradient clipping (by value) in the optimizer.
discriminator_optimizer = keras.optimizers.RMSprop(lr=0.0008,
clipvalue=1.0,
decay=1e-8)
discriminator.compile(optimizer=discriminator_optimizer,
loss='binary_crossentropy')
def retain(ARGS):
'''Create the model'''
#Define the constant for model saving
reshape_size = ARGS.emb_size + ARGS.numeric_size
if ARGS.allow_negative:
embeddings_constraint = FreezePadding()
beta_activation = 'tanh'
output_constraint = None
else:
embeddings_constraint = FreezePadding_Non_Negative()
beta_activation = 'sigmoid'
output_constraint = non_neg()
#Get available gpus , returns empty list if none
glist = get_available_gpus()
def reshape(data):
'''Reshape the context vectors to 3D vector'''
return K.reshape(x=data, shape=(K.shape(data)[0], 1, reshape_size))
#Code Input
codes = L.Input((None, None), name='codes_input')
inputs_list = [codes]
#Calculate embedding for each code and sum them to a visit level
codes_embs_total = L.Embedding(
ARGS.num_codes + 1,
ARGS.emb_size,
name='embedding',
embeddings_constraint=embeddings_constraint)(codes)
codes_embs = L.Lambda(lambda x: K.sum(x, axis=2))(codes_embs_total)
#Numeric input if needed
if ARGS.numeric_size:
numerics = L.Input((None, ARGS.numeric_size), name='numeric_input')
inputs_list.append(numerics)
full_embs = L.concatenate([codes_embs, numerics], name='catInp')
else:
full_embs = codes_embs
#Apply dropout on inputs
full_embs = L.Dropout(ARGS.dropout_input)(full_embs)
#Time input if needed
if ARGS.use_time:
time = L.Input((None, 1), name='time_input')
inputs_list.append(time)
time_embs = L.concatenate([full_embs, time], name='catInp2')
else:
time_embs = full_embs
#Setup Layers
#This implementation uses Bidirectional LSTM instead of reverse order
# (see https://github.com/mp2893/retain/issues/3 for more details)
#If training on GPU and Tensorflow use CuDNNLSTM for much faster training
if glist:
alpha = L.Bidirectional(L.CuDNNLSTM(ARGS.recurrent_size,
return_sequences=True),
name='alpha')
beta = L.Bidirectional(L.CuDNNLSTM(ARGS.recurrent_size,
return_sequences=True),
name='beta')
else:
alpha = L.Bidirectional(L.LSTM(ARGS.recurrent_size,
return_sequences=True,
implementation=2),
name='alpha')
beta = L.Bidirectional(L.LSTM(ARGS.recurrent_size,
return_sequences=True,
implementation=2),
name='beta')
alpha_dense = L.Dense(1, kernel_regularizer=l2(ARGS.l2))
beta_dense = L.Dense(ARGS.emb_size + ARGS.numeric_size,
activation=beta_activation,
kernel_regularizer=l2(ARGS.l2))
#Compute alpha, visit attention
alpha_out = alpha(time_embs)
alpha_out = L.TimeDistributed(alpha_dense,
name='alpha_dense_0')(alpha_out)
alpha_out = L.Softmax(axis=1)(alpha_out)
#Compute beta, codes attention
beta_out = beta(time_embs)
beta_out = L.TimeDistributed(beta_dense, name='beta_dense_0')(beta_out)
#Compute context vector based on attentions and embeddings
c_t = L.Multiply()([alpha_out, beta_out, full_embs])
c_t = L.Lambda(lambda x: K.sum(x, axis=1))(c_t)
#Reshape to 3d vector for consistency between Many to Many and Many to One implementations
contexts = L.Lambda(reshape)(c_t)
#Make a prediction
contexts = L.Dropout(ARGS.dropout_context)(contexts)
output_layer = L.Dense(1,
activation='sigmoid',
name='dOut',
kernel_regularizer=l2(ARGS.l2),
kernel_constraint=output_constraint)
#TimeDistributed is used for consistency
# between Many to Many and Many to One implementations
output = L.TimeDistributed(output_layer,
name='time_distributed_out')(contexts)
#Define the model with appropriate inputs
model = Model(inputs=inputs_list, outputs=[output])
return model
def Conv2DClassifierIn1(x_train,y_train,x_test,y_test):
summary = True
verbose = 1
# setHyperParams------------------------------------------------------------------------------------------------
batch_size = {{choice([512,256,128,64,32])}}
epoch = {{choice([300,275,250,225,200,175,150,125,100,75,50,25])}}
conv_block={{choice(['four', 'three', 'two'])}}
conv1_num={{choice([64,32,16,8])}}
conv2_num={{choice([128,64,32,16])}}
conv3_num={{choice([128,64,32])}}
conv4_num={{choice([256,128,64,32])}}
dense1_num={{choice([512, 256, 128])}}
dense2_num={{choice([256, 128, 64])}}
l1_regular_rate = {{uniform(0.00001, 1)}}
l2_regular_rate = {{uniform(0.000001, 1)}}
drop1_num={{uniform(0.1, 1)}}
drop2_num={{uniform(0.0001, 1)}}
activator={{choice(['tanh','relu','elu'])}}
optimizer={{choice(['SGD','rmsprop','adam'])}}
#---------------------------------------------------------------------------------------------------------------
kernel_size = (3, 3)
pool_size = (2, 2)
initializer = 'random_uniform'
padding_style = 'same'
loss_type='binary_crossentropy'
metrics=['accuracy']
my_callback = None
# early_stopping = EarlyStopping(monitor='val_loss', patience=4)
# checkpointer = ModelCheckpoint(filepath='keras_weights.hdf5',
# verbose=1,
# save_best_only=True)
# my_callback = callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.2,
# patience=5, min_lr=0.0001)
# build --------------------------------------------------------------------------------------------------------
input_layer = Input(shape=x_train.shape[1:])
conv = layers.Conv2D(conv1_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(input_layer)
conv = layers.Conv2D(conv1_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(conv)
pool = layers.MaxPooling2D(pool_size,padding=padding_style)(conv)
if conv_block == 'two':
conv = layers.Conv2D(conv2_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(pool)
conv = layers.Conv2D(conv2_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(conv)
BatchNorm = layers.BatchNormalization(axis=-1)(conv)
pool = layers.MaxPooling2D(pool_size,padding=padding_style)(BatchNorm)
elif conv_block == 'three':
conv = layers.Conv2D(conv2_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(pool)
conv = layers.Conv2D(conv2_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(conv)
BatchNorm = layers.BatchNormalization(axis=-1)(conv)
pool = layers.MaxPooling2D(pool_size,padding=padding_style)(BatchNorm)
conv = layers.Conv2D(conv3_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(pool)
conv = layers.Conv2D(conv3_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(conv)
BatchNorm = layers.BatchNormalization(axis=-1)(conv)
pool = layers.MaxPooling2D(pool_size,padding=padding_style)(BatchNorm)
elif conv_block == 'four':
conv = layers.Conv2D(conv2_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(pool)
conv = layers.Conv2D(conv2_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(conv)
BatchNorm = layers.BatchNormalization(axis=-1)(conv)
pool = layers.MaxPooling2D(pool_size,padding=padding_style)(BatchNorm)
conv = layers.Conv2D(conv3_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(pool)
conv = layers.Conv2D(conv3_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(conv)
BatchNorm = layers.BatchNormalization(axis=-1)(conv)
pool = layers.MaxPooling2D(pool_size,padding=padding_style)(BatchNorm)
conv = layers.Conv2D(conv4_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(pool)
conv = layers.Conv2D(conv4_num,kernel_size,padding=padding_style,kernel_initializer=initializer,activation=activator)(conv)
BatchNorm = layers.BatchNormalization(axis=-1)(conv)
pool = layers.MaxPooling2D(pool_size,padding=padding_style)(BatchNorm)
flat = layers.Flatten()(pool)
drop = layers.Dropout(drop1_num)(flat)
dense = layers.Dense(dense1_num, activation=activator, kernel_regularizer=regularizers.l1_l2(l1=l1_regular_rate,l2=l2_regular_rate))(drop)
BatchNorm = layers.BatchNormalization(axis=-1)(dense)
drop = layers.Dropout(drop2_num)(BatchNorm)
dense = layers.Dense(dense2_num, activation=activator, kernel_regularizer=regularizers.l1_l2(l1=l1_regular_rate,l2=l2_regular_rate))(drop)
output_layer = layers.Dense(len(np.unique(y_train)),activation='softmax')(dense)
model = models.Model(inputs=input_layer, outputs=output_layer)
if summary:
model.summary()
# train(self):
class_weights = class_weight.compute_class_weight('balanced', np.unique(y_train), y_train.reshape(-1))
class_weights_dict = dict(enumerate(class_weights))
model.compile(optimizer=optimizer,
loss=loss_type,
metrics=metrics # accuracy
)
result = model.fit(x=x_train,
y=y_train,
batch_size=batch_size,
epochs=epoch,
verbose=verbose,
callbacks=my_callback,
validation_data=(x_test, y_test),
shuffle=True,
class_weight=class_weights_dict
)
validation_acc = np.amax(result.history['val_acc'])
print('Best validation acc of epoch:', validation_acc)
return {'loss': -validation_acc, 'status': STATUS_OK, 'model': model}
# 计算val_steps，这是计算需要从val_gen中抽取多少次
val_steps=(valnum-lookback)//batch_size
# 查看训练集需要抽取的次数
train_steps=trainnum//batch_size
from keras.callbacks import EarlyStopping
early_stopping=EarlyStopping(monitor="val_loss",patience=30)
#创建模型
model1=models.Sequential()
model1.add(layers.Dense(512,activation="relu",input_shape=(lookback//step,data.shape[-1])))
model1.add(layers.Conv1D(filters=512,kernel_size=3,activation="relu"))
model1.add(layers.MaxPooling1D(5))
model1.add(layers.Conv1D(filters=512,kernel_size=3,activation="relu"))
model1.add(layers.GlobalMaxPool1D())
model1.add(layers.Dropout(0.5))
model1.add(layers.Dense(8,activation="softmax"))
model1.summary()
model1.compile(optimizer=optimizers.RMSprop(),
loss="categorical_crossentropy",
metrics=["acc"])
history1=model1.fit_generator(train_gen,
steps_per_epoch=train_steps,
epochs=200,
validation_data=val_gen,
validation_steps=val_steps,
callbacks=[early_stopping])
# 绘制结果
loss=history1.history["loss"]
