Basic usage of TensorFlow (v)--create neural networks and train

Article Author: Tyan
Blog: noahsnail.com |  CSDN | Pinterest

This article is mainly about the use of TensorFlow to create a simple neural network and training.

#!/usr/bin/env python # _*_ coding:utf-8 _*_ import tensorflow as TF import numpy as NP # Create a neural network layer def add_layer (input , In_size, out_size, activation_function = None): "" ":p Aram Input: Inputs:p The neural network layer Aram In_zize: output The size of the data:p Aram Out_size: The size of the output data:p Aram activation_function: Neural network activation function, default No "" "# defines the initial neural network Weight Weights = tf. Variable (Tf.random_normal ([In_size, Out_size]) # Defines the bias of the neural network biases = tf. Variable (Tf.zeros ([1, out_size]) + 0.1) # calculation W*x+b w_mul_x_plus_b = Tf.matmul (input, Weights) + biases # based on whether there is a stress Live functions are processed if activation_function is none:output = w_mul_x_plus_b Else:output = activation_function (w_mul_x_plus_b) Return output # Creates a three-layer neural network with input layer, hidden layer, output layer, the number of neurons is 1,10,1 # to create only one characteristic of the input data, the number of data is 300, the input layer X_data = Np.li Nspace ( -1, 1, +) [:, Np.newaxis] # Create noise in data noise = np.random.normal (0, 0.05, x_data.shape) # Create output for input data Y_data = Np.squ Is (X_data) + 1 + Noise # Defines the input data, none is the number of samples, the tableShows how much input data is OK, 1 is the number of features of the input data xs = tf.placeholder (Tf.float32, [None, 1]) # define output data, with xs equals ys = Tf.placeholder (Tf.float32, [None, 1]) # define a hidden layer Hidden_layer = Add_layer (xs, 1, ten, activation_function = tf.nn.relu) # define output Layer prediction = Add_layer (hidden_la Yer, 1, activation_function = None) # Solve the neural network parameters # define the loss function loss = Tf.reduce_mean (Tf.reduce_sum (Tf.square (ys-prediction ), reduction_indices = [1]) # defines the training process Train_step = Tf.train.GradientDescentOptimizer (0.1). Minimize (loss) # variable initialization init = TF . Global_variables_initializer () # defines session Sess = TF.  Session () # performs initialization work sess.run (INIT) # Training for I in range (1000): # Perform training and pass in data Sess.run (train_step, feed_dict = {xs):
X_data, ys:y_data}) if I% = = 0:print Sess.run (loss, feed_dict = {xs:x_data, ys:y_data}) # Close session Sess.close ()

The results of the implementation are as follows:

1.06731
0.0111914
0.00651229
0.00530187
0.00472237
0.00429948
0.00399815
0.00377548
0.00359714
0.00345819

ReferencesHttps://www.youtube.com/user/MorvanZhou