全连接层Layer

开启掘金成长之旅！这是我参与「掘金日新计划 · 12 月更文挑战」的第四天，点击查看活动详情

总结：此文为12月更文计划第四天第九篇。

全连接层Layer

layer是输入为5，输出为100的全连接层，所以对于输入为10x5的矩阵来看，会乘以一个5x100的矩阵，所以输出就是10x100. 首先导入这次所需要的包：

%matplotlib inline
import numpy as np
import sklearn
import pandas as pd
import os
import sys
import time
import tensorflow as tf

from tensorflow import keras

print(tf.__version__)
print(sys.version_info)
for module in mpl, np, pd, sklearn, tf, keras:
    print(module.__name__, module.__version__)

创建一个全连接层进行学习：

input_shape我们往往第一层指定输出就是10x100.

print(type(layer))
print('-'*50)
layer(tf.ones([10, 5]))  #这里是对应层的输出

对应的输出结果如下：

layer.variables 可以打印layer里包含的所有参数

# x * w + b  w就是指层的参数，kernel就是w，b就是bias
print(layer.variables)
print('-'*50)
layer.trainable_variables

#获得所有可训练的变量,和上面的变量数目一致 输出的结果如下：

对layer进行实战学习：

from sklearn.datasets import fetch_california_housing

housing = fetch_california_housing()
print(housing.DESCR)
print(housing.data.shape)
print(housing.target.shape)

导入sklearn数据包：

对这些数据进行学习，划分训练集与测试集

from sklearn.model_selection import train_test_split

x_train_all, x_test, y_train_all, y_test = train_test_split(
    housing.data, housing.target, random_state = 7)
x_train, x_valid, y_train, y_valid = train_test_split(
    x_train_all, y_train_all, random_state = 11)
print(x_train.shape, y_train.shape)
print(x_valid.shape, y_valid.shape)
print(x_test.shape, y_test.shape)

class CustomizedDenseLayer(keras.layers.Layer):
    def __init__(self, units, activation=None, **kwargs):
        self.units = units
        self.activation = keras.layers.Activation(activation)  #直接使用tf提供的
        super(CustomizedDenseLayer, self).__init__(**kwargs)
    
    def build(self, input_shape):
        """构建所需要的参数，也就是kernel还有bias"""
        # x * w + b. input_shape:[None, a] w:[a,b]output_shape: [None, b]
        print('-'*50)
        print(input_shape)
        self.kernel = self.add_weight(name = 'kernel',
                                      shape = (input_shape[1], self.units),
                                      initializer = 'uniform',#使用均匀分布的方法去初始化kernel
                                      trainable = True)
        self.bias = self.add_weight(name = 'bias',
                                    shape = (self.units, ),
                                    initializer = 'zeros',
                                    trainable = True)
        #接着我们要继承父类的build
        super(CustomizedDenseLayer, self).build(input_shape)

对神经网络进行正向学习：

model = keras.models.Sequential([
    CustomizedDenseLayer(30, activation='relu',
                         input_shape=x_train.shape[1:]),  #这里传入的是特征数
#     CustomizedDenseLayer(1),
    #再加一个激活函数层，这个和下面注释的两行等价的
#     customized_softplus,
    CustomizedDenseLayer(1,activation=customized_softplus),
#     keras.layers.Dense(1, activation="softplus"),  #一层
#     keras.layers.Dense(1), keras.layers.Activation('softplus'), #两层
])

输出的结果如下：