Operations on random variables not working properly in Tensorflow

Question

I create two Tensors (namely: x1, y2) which initialized with uniform distribution, But when I print out the result they were not what I expected.

This is my code:

x1 = tf.random_uniform([1], 0, 10, tf.int32)
y1 = tf.random_uniform([1], 0, 10, tf.int32)

subtraction = x1 - y1

with tf.Session() as sess:

    print(sess.run(x1))
    print(sess.run(y1))
    print(sess.run(subtraction))

This is the result:

[6]

[2]

[0]

Answer 1

In your code, x1 and y1 are random number generators. They take different values each time they are called. So when you call subtraction , which in turns call your number generators x1 and y1 , there is no reason to obtain results that are consistent with previous calls.

To achieve what you are looking for, store the values in a Variable :

import tensorflow as tf

x1 = tf.Variable(tf.random_uniform([1], 0, 10, tf.int32))
y1 = tf.Variable(tf.random_uniform([1], 0, 10, tf.int32))

subtraction = x1 - y1

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    print(sess.run(x1))
    print(sess.run(y1))
    print(sess.run(subtraction))

Alternatively, if you don't need persistence between iterations and can call all the operators relying on your number generators at once, pack them into the same call to sess.run :

import tensorflow as tf

x1 = tf.random_uniform([1], 0, 10, tf.int32)
y1 = tf.random_uniform([1], 0, 10, tf.int32)

subtraction = x1 - y1

with tf.Session() as sess:
    print(sess.run([x1, y1, subtraction]))