머신러닝스터디/2016/2016_05_28

import tensorflow as tf
# AND          OR           NXOR          XOR
# (0, 0) => 0  (0, 0) => 0  (0, 0) => 1  (0, 0) => 0
# (0, 1) => 0  (0, 1) => 1  (0, 1) => 0  (0, 1) => 1
# (1, 0) => 0  (1, 0) => 1  (1, 0) => 0  (1, 0) => 1
# (1, 1) => 1  (1, 1) => 1  (1, 1) => 1  (1, 1) => 0

W1 = tf.Variable(tf.random_uniform([2, 2]))
b1 = tf.Variable(tf.random_uniform([2]))

W2 = tf.Variable(tf.random_uniform([2, 1]))
b2 = tf.Variable(tf.random_uniform([1]))

def logic_gate(x):
    hidden = tf.sigmoid(tf.matmul(x, W1) + b1)
    return tf.sigmoid(tf.matmul(hidden, W2) + b2)

x = tf.placeholder("float", [None, 2])
y = tf.placeholder("float", [None, 1])

value = logic_gate(x)
loss = tf.reduce_sum(tf.pow(y-value, 2))
# loss = tf.reduce_mean(-tf.reduce_sum(y*tf.log(value), reduction_indices=1)) # Don't work
optimize = tf.train.GradientDescentOptimizer(0.01).minimize(loss)

init = tf.initialize_all_variables()

with tf.Session() as sess:
    sess.run(init)
    for i in range(30001):
        result = sess.run(optimize, feed_dict={x: [[0, 0], [0, 1], [1, 0], [1, 1]], y: [[1], [0], [0], [1]]})
        if (i % 1000 == 0):
            print("Epoch: ", i)
            print(sess.run([value, loss], feed_dict={x: [[0, 0], [0, 1], [1, 0], [1, 1]], y: [[1], [0], [0], [1]]}))