In [1]:

    

import tensorflow as tf


    

In [2]:

    

g = tf.Graph()


    

In [3]:

    

with g.as_default():
    X = tf.placeholder(tf.float32, [4, 2], name='X')
    Y = tf.placeholder(tf.float32, [4, 1], name='Y')
    W1 = tf.Variable(tf.random_uniform([2,2], -1, 1), name='w1')
    b1 = tf.Variable(tf.zeros([2]), name='b1')
    W2 = tf.Variable(tf.random_uniform([2, 1], -1, 1), name='w2')
    b2 = tf.Variable(tf.zeros([1]), name='b2')
    layer_one = tf.sigmoid(tf.matmul(X, W1) + b1)
    pred = tf.sigmoid(tf.matmul(layer_one, W2) + b2)
    cost = tf.reduce_mean(((Y * tf.log(pred)) + ((1 - Y) * tf.log(1.0 - pred))) * -1)
    train_step = tf.train.GradientDescentOptimizer(.01).minimize(cost)
    init = tf.initialize_all_variables()


    

In [4]:

    

sess = tf.InteractiveSession(graph=g)


    

In [5]:

    

sess.run(init)


    

In [6]:

    

import numpy as np


    

In [7]:

    

data = {X: np.array([[0,0], [0,1], [1, 0], [1,1]]).reshape(4, 2), Y:np.array([False, True, True, False]).reshape(4, 1)}


    

In [8]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [9]:

    

print(sess.run(cost, feed_dict=data))


    

In [10]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [11]:

    

print(sess.run(cost, feed_dict=data))


    

In [12]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [13]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [14]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [15]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [16]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [17]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [18]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [19]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [20]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [21]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [22]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [23]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [24]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [25]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [26]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [27]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [28]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [29]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [30]:

    

for i in range(10000):
    sess.run(train_step, feed_dict=data)


    

In [31]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [32]:

    

for i in range(50000):
    sess.run(train_step, feed_dict=data)


    

In [33]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [34]:

    

for i in range(50000):
    sess.run(train_step, feed_dict=data)


    

In [35]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [36]:

    

for i in range(50000):
    sess.run(train_step, feed_dict=data)


    

In [37]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [38]:

    

print(sess.run(W1))


    

In [39]:

    

for i in range(50000):
    sess.run(train_step, feed_dict=data)


    

In [40]:

    

print(sess.run(W1))


    

In [41]:

    

print(1-sess.run(cost, feed_dict=data))


    

In [42]:

    

%notebook -e example.ipynb