Sigmoid Activation Function: Understand It Visually and with Python

The Sigmoid activation function is one of the most important functions to understand when beginning machine learning and neural networks.

Its job is simple:

Sigmoid converts any real number into a value between 0 and 1.

Mathematically:

sigmoid(z) = 1 / (1 + e-z)

This makes Sigmoid particularly useful when the output of a model needs to represent something similar to a probability.

What Happens to Different Inputs?

Consider a few values:

Input z Sigmoid Output Meaning
Large negative Close to 0 Strongly towards Class 0
0 0.5 Decision boundary
Large positive Close to 1 Strongly towards Class 1

1. Interactive Sigmoid Visualizer

Start with the interactive version below.

Move the slider or enter different values of z. Watch how the point moves along the Sigmoid curve and how the corresponding probability changes.

Try These Values

Experiment with:

  • z = -10
  • z = -5
  • z = -1
  • z = 0
  • z = 1
  • z = 5
  • z = 10

Notice an important property:

When z = 0, Sigmoid returns exactly 0.5.

2. Now Implement the Same Idea in Python

Once you understand the curve visually, run the Python program below.

This version uses:

  • NumPy to calculate the Sigmoid function.
  • Matplotlib to plot the S-shaped curve.
  • A user-entered value of z.
  • A threshold of 0.5 to demonstrate binary classification.

Understanding the Python Program

The central function is: 

def sigmoid(z):
    return 1 / (1 + np.exp(-z))

NumPy calculates e-z, and the result is placed into the Sigmoid formula.

The program then converts the output into a percentage and applies a simple decision rule: 

if probability >= 0.5:
    print("CLASS 1")
else:
    print("CLASS 0")

Why Use 0.5?

The Sigmoid value at z = 0 is:

sigmoid(0) = 0.5

Therefore, 0.5 is a convenient example of a decision threshold.

In a real machine-learning application, however, the threshold does not always have to be 0.5. It can be adjusted according to the problem.

From a Number to a Classification

The complete idea can now be seen as a sequence:

Input z → Sigmoid → Value between 0 and 1 → Threshold → Class

For example:

z Approx. Sigmoid Threshold 0.5
-5 0.0067 Class 0
-1 0.2689 Class 0
0 0.5000 Class 1
1 0.7311 Class 1
5 0.9933 Class 1

The Most Important Idea

Do not think of Sigmoid as something mysterious that exists only inside a neural network.

At its core, it is simply a mathematical transformation:

Give Sigmoid any number — negative, zero or positive — and it returns a smoothly scaled value between 0 and 1.

Experiment with both programs above. Change the value of z, predict the result first, and then compare your prediction with the actual Sigmoid output.

Learn With Champak
Learn programming, artificial intelligence and machine learning by running the programs yourself.