Tensor math operations in TensorFlow

Introduction

Tensors are like multi-dimensional arrays. Doing math with tensors helps computers learn patterns and solve problems.

When you want to add or multiply data in machine learning models.

When you need to calculate predictions from input data.

When you want to combine or transform data in neural networks.

When you want to compute loss or accuracy during training.

When you want to apply mathematical functions like square root or exponent.

Syntax

TensorFlow

import tensorflow as tf

# Basic math operations
result = tf.add(tensor1, tensor2)  # addition
result = tf.subtract(tensor1, tensor2)  # subtraction
result = tf.multiply(tensor1, tensor2)  # element-wise multiplication
result = tf.divide(tensor1, tensor2)  # element-wise division

# Other operations
result = tf.matmul(tensor1, tensor2)  # matrix multiplication
result = tf.sqrt(tensor)  # square root
result = tf.exp(tensor)  # exponentiation

Use tf.add, tf.subtract, tf.multiply, and tf.divide for element-wise math.

Use tf.matmul for matrix multiplication, which is different from element-wise multiplication.

Examples

Adds two 1D tensors element-wise: [1+4, 2+5, 3+6] = [5, 7, 9]

TensorFlow

a = tf.constant([1, 2, 3])
b = tf.constant([4, 5, 6])
sum_ab = tf.add(a, b)

Performs matrix multiplication of two 2x2 tensors.

TensorFlow

a = tf.constant([[1, 2], [3, 4]])
b = tf.constant([[5, 6], [7, 8]])
product = tf.matmul(a, b)

Computes square root of each element: [3.0, 4.0, 5.0]

TensorFlow

a = tf.constant([9.0, 16.0, 25.0])
sqrt_a = tf.sqrt(a)

Sample Model

This program shows basic tensor math: addition, element-wise multiplication, matrix multiplication, and square root.

TensorFlow

import tensorflow as tf

# Define two tensors
x = tf.constant([[1, 2], [3, 4]], dtype=tf.float32)
y = tf.constant([[5, 6], [7, 8]], dtype=tf.float32)

# Element-wise addition
add_result = tf.add(x, y)

# Element-wise multiplication
mul_result = tf.multiply(x, y)

# Matrix multiplication
matmul_result = tf.matmul(x, y)

# Square root of elements in x
sqrt_x = tf.sqrt(x)

# Print results
print("Addition:\n", add_result.numpy())
print("Element-wise Multiplication:\n", mul_result.numpy())
print("Matrix Multiplication:\n", matmul_result.numpy())
print("Square Root of x:\n", sqrt_x.numpy())

Important Notes

Tensors must have compatible shapes for operations like addition or multiplication.

Matrix multiplication requires the inner dimensions to match (e.g., (2,3) x (3,4)).

Use .numpy() to convert tensors to arrays for easy printing.

Summary

Tensors are multi-dimensional arrays used in machine learning.

Tensor math operations include element-wise and matrix multiplication.

TensorFlow provides simple functions like tf.add and tf.matmul to do these operations.