Continuing with the Tensorflow series, there are the following articles:

  1. Hello world from TensorFlow
  2. Basic math with TensorFlow
  3. Handling matrices with Tensorflow

The Tensorflow Variable object lets you define a variable (store a state in a Graph object with an initial value), whereas the placeholder object is used to feed external data into a Graph object, but the data to be fed is not supplied when the placeholder is created; for that, the initialization of the variables is used.

Below is the example of using variables and placeholder from jupyter notebook:

In [1]:

#Se importa tensofflow
import tensorflow as tf

In [2]:

#Se crea una sesión interactiva ( útil cuando se usa jupyter notebook)
sess = tf.InteractiveSession()

In [3]:

#Se crea tensor como un número uniforme aleatorio, una matrix de 4x4, de elementos
#de valores entre 0 a 1.
tensor = tf.random_uniform((4,4),0,1)

In [4]:

#tensor es un objeto Tensor, de números aleatorios uniformes, de una matrix 4x4 y de typo float32.
tensor

Out[4]:

<tf.Tensor 'random_uniform:0' shape=(4, 4) dtype=float32>

In [5]:

#se crea variable como un objeto variable que se le pasa como valor inicial el objeto tensor.
variable = tf.Variable(initial_value=tensor)

In [6]:

#Se imprime el valor de la variable, 
#esta devuelve tipo Variable, con una matriz 4x4 con elementos de tipo float32
print(variable)

<tf.Variable 'Variable:0' shape=(4, 4) dtype=float32_ref>

In [7]:

#Se inicializa las variables con init
init = tf.global_variables_initializer()

In [8]:

#Se corre la inicialización
sess.run(init)

In [9]:

#Se muestra el valor de la variable corriendo la sesión, se nota los elementos
#entre 0 a 1 y de tipo float32
sess.run(variable)

Out[9]:

array([[0.05263364, 0.94140494, 0.4843645 , 0.36463213],
       [0.9149779 , 0.01328921, 0.3837136 , 0.7204586 ],
       [0.5955255 , 0.70812213, 0.19933975, 0.40913558],
       [0.359192  , 0.82358146, 0.1225971 , 0.24963093]], dtype=float32)

In [10]:

#Se crea el placeholder, lo cual reserva el espacio, pero no tiene datos como una variable.
ph = tf.placeholder(tf.float32,shape=(None,5))

In [11]:

#Se muestra el objeto placeholder, sólo tiene definido el tipo de datos el cual es float32
print(ph)

Tensor("Placeholder:0", shape=(?, 5), dtype=float32)

In [ ]:

undefined

The next article will explain the use of Graph.

Make a donation! If you liked the article you can make a donation with Bitcoin (BTC) using the digital wallet of your choice at the following address: 17MtNybhdkA9GV3UNS6BTwPcuhjXoPrSzV

Or by scanning the QR code from your wallet:

17MtNybhdkA9GV3UNS6BTwPcuhjXoPrSzV