【Keras学习】构建神经网络的两种写法Sequential及layer类后加(layer)

最近在读paper，看keras复现模型时看到了除了用Sequential()不断add(layers)的写法之外还有另一种方式，即在layer类后加(layer)这种方式，花了些时间去理解和学习，在此分享一下两种表述。

以下两种代码搭建的是相同的网络结构。
[1] 利用Sequential()作为起始，不断add()后面的layer，架构线性表示，即
$y=f(x)\to z=g(y)\to m=h(z)\to \dots$

from keras.models import Sequential
from keras.layers import Conv2d, Maxpooling2D
from keras.layers import Dense, Dropout, Flatten

model = Sequential()

# input: 100x100 images with 3 channels -> (100, 100, 3) tensors.
# this applies 32 convolution filters of size 3x3 each.

model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(100, 100, 3)))
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))

model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))

model.add(Flatten())
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(10, activation='softmax'))

sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy',
              optimizer=sgd)

[2] 采用一种layer类后加“(layer)”的方式进行搭建，理解上更接近嵌套式：
$f(x)\to g(f(x))\to h(g(f(x)))$
（以下未经程序验证是否可行，仅分享表达方式的理解）

from keras.models import Sequential
from keras.layers import Conv2d, Maxpooling2D
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Input, Output
from keras.models import Model

inputs = Input(shape=(100, 100, 3))
layer = Conv2d(32, (3, 3), activation='relu')(inputs)
# 便于理解，这种表示可以分解为两步
# convlayer = Conv2d(32, (3, 3), activation='relu')
# layer = convlayer(inputs)
layer = Conv2d(32, (3, 3), activation='relu')(layer)
layer = Maxpooling2D(pool_size=(2, 2))(layer)
layer = Dropout(0.25)(layer)

layer = Conv2d(64, (3, 3), activation='relu')(layer)
layer = Conv2d(64, (3, 3), activation='relu')(layer)
layer = Maxpooling2D(pool_size=(2, 2))(layer)
layer = Dropout(0.25)(layer)

layer = Flatten()(layer)
layer = Dense(256, activation='relu')(layer)
layer = Dropout(0.5)(layer)
outputs = Dense(10, activation='softmax')(layer)
model = Model(inputs=[inputs], outputs=[outputs])

# compile步骤相同
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy',
              optimizer=sgd)