Published July 22, 2022 #MachineLearning

Understanding Softmax Activation Function: A Key Element in Machine Learning

The softmax activation function is a crucial element in the field of machine learning, particularly in the realm of multi-class classification problems. It plays a vital role in converting raw scores or logits into probabilities, enabling us to make informed decisions and predictions. In this article, we will dive deep into the concept of softmax, its mathematical representation, and how to implement it using Python.

The Concept of Softmax

In the context of machine learning, softmax is used to transform a vector of real numbers into a probability distribution. Given a set of values, it assigns probabilities to each element such that the sum of all probabilities is equal to one. The softmax function is especially useful when dealing with multi-class classification tasks, where we need to categorize input data into one of several possible classes.

Mathematical Representation

The softmax function takes a vector of real numbers as input and returns a vector of probabilities. Given a vector of logits (raw scores) denoted as $Z = [Z_1, Z_2, …, Z_n]$, the softmax function for element $Z_i$ can be mathematically represented as follows:

$$ \text{Softmax}(Z_i) = \frac{e^{Z_i}}{\sum_{j=1}^{n} e^{Z_j}} $$


  • $e$ is the base of the natural logarithm (Euler’s number).
  • $Z_i$ is the $i$-th element of the input vector $Z$.
  • The denominator is the sum of the exponential values of all elements in $Z$.

The softmax function exponentiates each element of the input vector, making them positive, and then normalizes them to obtain probabilities.

Python Implementation

Now, let’s implement the softmax function in Python:

import numpy as np

def softmax(Z):
    exp_Z = np.exp(Z)
    sum_exp_Z = np.sum(exp_Z)
    softmax_probs = exp_Z / sum_exp_Z
    return softmax_probs

In this Python code, we use the NumPy library to efficiently compute the softmax probabilities. The function takes an array $Z$ as input and returns the corresponding probabilities.

Example Usage

Let’s see an example of using the softmax function:

# Input logits (raw scores)
logits = np.array([2.0, 1.0, 0.1])

# Applying softmax
probabilities = softmax(logits)



[0.65900114 0.24243297 0.09856589]

The output array represents the probabilities of each class. In this example, class 1 has the highest probability of approximately 66%, followed by class 2 with around 24%, and class 3 with around 9%.


In conclusion, the softmax activation function is a fundamental tool in machine learning, particularly for multi-class classification tasks. It allows us to convert raw scores into meaningful probabilities, facilitating better decision-making and prediction in various applications. By understanding the concept of softmax and implementing it in Python, we can leverage its power to tackle a wide range of classification problems.

Next Post: Mlflow