Tensorflow如何打印神经网络权重参数

在Tensorflow中，神经网络的权重是一个variable，所以要打印出来只需要按下面的代码即可。最重要的就是sess.run()

import tensorflow as tf

weight = tf.Variable(tf.random_normal([1,2]))
sess = tf.Session()
init = tf.global_variables_initializer()
sess.run(init)
print(sess.run(weight))

但是如果使用的一个函数来添加隐藏层和输出层，权重的变量初始化在函数里面，就需要将权重作为结果返回。举个?：

下面建立一个神经网络，用add_layer()函数来添加隐藏层和输出层。我们可以看到最后会返回一个weights的list。

import numpy as np
import tensorflow as tf

ACTIVATION = tf.nn.relu
N_LAYERS = 3
N_HIDDEN_UNITS = 15

下面这个是一个种子函数，让每一次的运行结果都保持一致。 

def fix_seed(seed=1):
    # reproducibel
    np.random.seed(seed)
    tf.set_random_seed(seed)

 神经网络的结构函数，里面包含了添加隐藏层的函数add_player()：

def built_net(xs, ys):
    def add_layer(inputs, in_size, out_size, activation_function=None):
        # weights and biases
        Weights = tf.Variable(tf.random_normal([in_size, out_size], mean=0., stddev=1.))
        biases = tf.Variable(tf.zeros([1, out_size])+0.1)
#         w = Weights.read_value()
        # fully connected product
        Wx_plus_b = tf.matmul(inputs, Weights)+biases
 
        # activation
        if activation_function is None:
            outputs = Wx_plus_b
        else:
            outputs = activation_function(Wx_plus_b)
            
        return outputs, Weights
    
    fix_seed(1)
    
    # record inputs for every layer
    layers_inputs = [xs]
    w_outs = []
    
    # build hidden layers
    for l_n in range(N_LAYERS):
        layer_input = layers_inputs[l_n]
        in_size = layers_inputs[l_n].get_shape()[1].value
        output, weights= add_layer(
            layer_input,
            in_size,
            N_HIDDEN_UNITS,
            ACTIVATION,
            norm,
        )
        layers_inputs.append(output)
        w_outs.append(weights)
    
    ##output layer 
    prediction, weights = add_layer(layers_inputs[-1], 15, 1,
                           activation_function=None)
    
    cost = tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction), reduction_indices=[1]))
    train_op = tf.train.GradientDescentOptimizer(0.001).minimize(cost)
    return [train_op, cost, layers_inputs, w_outs]

 建立要用的数据集

# make up data
fix_seed(1)
x_data = np.linspace(-7, 10, 2500)[:, np.newaxis]
np.random.shuffle(x_data)
noise = np.random.normal(0, 8, x_data.shape)
y_data = np.square(x_data) - 5 + noise

运行网络，并打印出weights变量 

xs = tf.placeholder(tf.float32, [None, 1])#[num_samples, num_features]
ys = tf.placeholder(tf.float32, [None, 1])

train_op, cost, layers_inputs, weights = built_net(xs, ys)

sess = tf.Session()

init = tf.global_variables_initializer()
sess.run(init)

for i in range(251):
    if i%50 == 0:
        print("------------")
        print(sess.run(weights))
    sess.run(train_op, feed_dict={xs:x_data[i*10:i*10+10], ys:y_data[i*10:i*10+10]})

参考连接：https://morvanzhou.github.io/tutorials/machine-learning/tensorflow/5-13-BN/