tensorflow 入門級别程式實作

import tensorflow as tf
from numpy.random import RandomState
import math
learning_rate = 0.01
MAX_STEPS = 1001
rdm = RandomState(1)
dataset_size = 256
validation_rate = 0.2

validation_size = math.floor(dataset_size * validation_rate)

X = rdm.rand(dataset_size, 2)
Y = [[int(x1 + x2 < 1)] for (x1, x2) in X]

train_X = X[validation_size:]
train_Y = Y[validation_size:]
validation_X = X[: validation_size]
validation_Y = Y[: validation_size]

x = tf.placeholder(tf.float32, shape=(None, 2), name="x_input")
y_ = tf.placeholder(tf.float32, shape=(None, 1), name="y_input")

input_weight = tf.Variable(tf.random_normal((2, 3)), name="input_weight")
input_bias = tf.Variable(tf.constant(0.1, shape=[3]), name="input_bias")

hidden_weight = tf.Variable(tf.random_normal((3, 3)), name="hidden_weight")
hidden_bias = tf.Variable(tf.constant(0.1, shape=[3,]), name="hidden_bias")

output_weight = tf.Variable(tf.random_normal((3, 1)), name="output_weight")
output_bias = tf.Variable(tf.constant(0.1, shape=[1]), name="output_bias")

#input layer
z = tf.matmul(x, input_weight) + input_bias
y = tf.sigmoid(z)
#hidden layer
z = tf.matmul(y, hidden_weight) + hidden_bias
y = tf.sigmoid(z)

#output layer
z = tf.matmul(y, output_weight) + output_bias
y = tf.sigmoid(z)

cross_entropy = - tf.reduce_mean(y_ * tf.log(tf.clip_by_value(y, 1e-3, 1.0))
                                 + (1-y_)*tf.log(tf.clip_by_value(1-y, 1e-3, 1.0)))
tf.summary.scalar("cross entropy", cross_entropy)

optimzer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cross_entropy)

merged = tf.summary.merge_all()

with tf.Session() as sess:
    summary_writer = tf.summary.FileWriter("./train", sess.graph)
    summary_writer1 = tf.summary.FileWriter("./test", sess.graph)
    tf.global_variables_initializer().run()
    for step in range(MAX_STEPS):
        summary_train, _ = sess.run([merged, optimzer], feed_dict={x: train_X, y_: train_Y})
        summary_writer.add_summary(summary_train, step)
        summary_test, _ = sess.run([merged, cross_entropy], feed_dict={x: validation_X, y_: validation_Y})
        summary_writer1.add_summary(summary_test, step)
    summary_writer.close()
    summary_writer1.close()
           

設計為: 實作三層的神經網絡，輸出端損失函數為cross entropy

實驗結果如下：

訓練1001次結果：

藍線表示validation  

紅線表示train

訓練10001次結果: 并沒有出現過拟合，我的了解可能是因為神經網絡學到了rand的屬性，是以才能拟合的很好