keras 构建神经网络流程

import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Activation
from keras import optimizers
from keras import losses
from keras import metrics
import keras.backend as K
from keras.callbacks import EarlyStopping
'''
    1. 准备数据
    2. 导入包
    3. 创建顺序模型
    4. add 网络层（第一层设定输入维度，最后一层设定输出维度，中间自动识别）
    5. 编译（优化器optimizer的选择，损失函数loss的选择，评价指标metrics的选择）
    6. 训练（fit，epochs，batch_size，callbacks）
    7. 预测（predict）
'''
X = np.random.random((1000, 100))
Y = np.random.randint(10, size=(1000, 1))


def mean_pred(y_true, y_pred):
    return K.mean(y_pred)


model = Sequential()
model.add(Dense(32, activation='relu', input_dim=784))
model.add(Dense(100, activation='softmax'))

model.compile(optimizer=optimizers.SGD(learning_rate=0.01, momentum=0.001), loss=losses.categorical_crossentropy, metrics=[metrics.accuracy, mean_pred])
early_stop = EarlyStopping(monitor='val_loss', patience=15, verbose=0)
model.fit(X, Y, epochs=100, batch_size=64, callbacks=[early_stop])

model.predict(X)


'''
    不同的构建方式
'''
from keras.layers import Input
from keras.models import Model

inputs = Input(shape=(784,))

dense_1 = Dense(32, activation='relu')(inputs)
dense_2 = Dense(64, activation='relu')(dense_1)
output = Dense(100, activation='softmax')(dense_2)

model = Model(inputs=inputs, outputs=output)
model.compile(optimizer=optimizers.SGD(learning_rate=0.01, momentum=0.001), loss=losses.categorical_crossentropy, metrics=[metrics.accuracy, mean_pred])
model.fit(X, Y, epochs=100, batch_size=64, callbacks=[early_stop])