Supervised Learning

Supervised learning is a machine learning method where models are trained using labeled data so it can predict correct outputs on new data.

Chandan KumarUpdated January 29, 2026Founder of Geekflare

At a glance

In 1957, Frank Rosenblatt developed the Perceptron, the first neural network for supervised pattern recognition, proving machines could learn from examples
In 1981, Stevo Bozinovski demonstrated a neural network using supervised learning to recognize 40 patterns, including letters, numbers, and symbols

ELI5 (Explain Like I'm 5)

Imagine you’re learning math with a teacher.

The teacher gives you a question and also tells you the correct answer. When you make a mistake, the teacher corrects you. Over time, you learn how to solve similar problems on your own.

That’s supervised learning.

The computer is the student, and the labeled examples are the teacher showing the right answers.

Once the computer has practiced enough, it can try answering new questions by itself—even ones it hasn’t seen before.

How does Supervised Learning Work?

In supervised learning, a model is first trained on labeled data.

Each training example includes:

an input (such as an image, text, or numerical data)
a label (the correct output)

The model learns a relationship between inputs and outputs so it can make predictions on unseen data.

The training involves these steps:

Collect and label training data
Feed the data into a learning algorithm
Compare predictions with known labels
Adjust internal parameters to reduce errors
Repeat until performance stabilizes

This feedback-driven loop lets the model improve over time.

Types of Supervised Learning

Most supervised learning problems fall into one of these 2 types.

Classification: which involves predicting categories (for example, spam vs not spam)
Regression: which is predicting numeric values (for example, stock prices or demand forecasts)