Constants and variables

Introduction to TensorFlow in Python

Isaiah Hull

Visiting Associate Professor of Finance, BI Norwegian Business School

What is TensorFlow?

Open-source library for graph-based numerical computation
- Developed by the Google Brain Team
Low and high level APIs
- Addition, multiplication, differentiation
- Machine learning models
Important changes in TensorFlow 2.0
- Eager execution by default
- Model building with Keras and Estimators

What is a tensor?

Generalization of vectors and matrices
Collection of numbers
Specific shape

What is a tensor?

The image shows a slice of bread, a slice of bread divided into 9 pieces, and a loaf of bread. Source: Public Domain Vectors

Defining tensors in TensorFlow

import tensorflow as tf

# 0D Tensor
d0 = tf.ones((1,))

# 1D Tensor
d1 = tf.ones((2,))

# 2D Tensor
d2 = tf.ones((2, 2))

# 3D Tensor
d3 = tf.ones((2, 2, 2))

Defining tensors in TensorFlow

# Print the 3D tensor
print(d3.numpy())

[[[1. 1.]
  [1. 1.]]

 [[1. 1.]
  [1. 1.]]]

Defining constants in TensorFlow

A constant is the simplest category of tensor
- Not trainable
- Can have any dimension

from tensorflow import constant

# Define a 2x3 constant.
a = constant(3, shape=[2, 3])

# Define a 2x2 constant.
b = constant([1, 2, 3, 4], shape=[2, 2])

Using convenience functions to define constants

Operation	Example
`tf.constant()`	`constant([1, 2, 3])`
`tf.zeros()`	`zeros([2, 2])`
`tf.zeros_like()`	`zeros_like(input_tensor)`
`tf.ones()`	`ones([2, 2])`
`tf.ones_like()`	`ones_like(input_tensor)`
`tf.fill()`	`fill([3, 3], 7)`

Defining and initializing variables

import tensorflow as tf

# Define a variable
a0 = tf.Variable([1, 2, 3, 4, 5, 6], dtype=tf.float32)
a1 = tf.Variable([1, 2, 3, 4, 5, 6], dtype=tf.int16)

# Define a constant
b = tf.constant(2, tf.float32)

# Compute their product
c0 = tf.multiply(a0, b)
c1 = a0*b

Let's practice!

Introduction to TensorFlow in Python