Skip to main content

​
What is Supervised Learning

Machine learning is creating tremendous economic value today. I think 99 percent of the economic value created by machine learning today is through one type of machine learning, which is called supervised learning.
Supervised machine learning refers to algorithms that learn x to y or input to output mappings. The key characteristic of supervised learning is that you give your learning algorithm examples to learn from, including the right answers.
By seeing correct pairs of input x and desired output label y, the learning algorithm eventually learns to take just the input alone without the output label and gives a reasonably accurate prediction or guess of the output.

​
Real-World Examples

Supervised learning powers many applications you use every day:
If the input x is an email and the output y is “spam” or “not spam”, this gives you your spam filter. Nearly every email service uses machine learning to filter spam effectively.
If the input is an audio clip and the algorithm’s job is to output the text transcript, then this is speech recognition. This technology powers voice assistants and transcription services.
The most lucrative form of supervised learning today is probably used in online advertising. Large online ad platforms use learning algorithms that input information about an ad and information about you, then predict if you will click on that ad. Every click generates revenue for these companies.
For autonomous vehicles, the learning algorithm takes as input an image and information from sensors like radar, then outputs the position of other cars so your self-driving car can safely navigate around them.

​
Types of Supervised Learning

There are two main types of supervised learning algorithms:

Regression

Regression algorithms predict numbers from infinitely many possible values. Examples include predicting housing prices (150,000or150,000 or 183,000) or temperature values.

Classification

Classification algorithms predict categories from a small finite set of possible outputs. Examples include determining if an email is spam/not spam, or if a tumor is malignant/benign.

​
Regression vs Classification

​
Regression Algorithm

Let’s look at predicting housing prices based on house size. You collect data and plot it:
  • Horizontal axis: Size of house in square feet
  • Vertical axis: Price in thousands of dollars
With this data, a learning algorithm can fit a straight line or curve to predict prices. For example, a 750 square foot house might be predicted to sell for 150,000withalinearmodel,or150,000 with a linear model, or 200,000 with a curved model.
Regression problems involve predicting a number from infinitely many possible numbers, such as house prices which could be any value like 150,000, 183,000, or any number in between.

​
Classification Algorithm

Classification predicts categories rather than continuous numbers. Consider breast cancer detection:
  • Malignant (cancerous): Class 1 or positive class
  • Benign (not cancerous): Class 0 or negative class
You can plot tumor size on the horizontal axis and the class (0 or 1) on the vertical axis. The algorithm learns to classify new tumors based on their features.
Classification problems can use multiple input features. For breast cancer detection, you might use tumor size, patient age, thickness of tumor clump, uniformity of cell size, and uniformity of cell shape to improve prediction accuracy.

​
Binary Classification

When there are only two possible outputs, this is called binary classification:
  • Negative class (0, false, no): Represents absence of the property
  • Positive class (1, true, yes): Represents presence of the property
Examples:
  • Email: not spam (0) or spam (1)
  • Transaction: legitimate (0) or fraudulent (1)
  • Tumor: benign (0) or malignant (1)

​
Key Concepts

1

Training with Examples

First, train your model with examples of inputs x and the correct answers (labels y).
2

Learning Patterns

The model learns from these input-output pairs to identify patterns and relationships.
3

Making Predictions

After training, the model can take a brand new input x that it has never seen before and produce an appropriate output y.

​
Conclusion

The two major types of supervised learning are regression and classification:
  • Regression: Predicts numbers from infinitely many possible output numbers (e.g., house prices)
  • Classification: Makes predictions of a category from a small set of possible outputs (e.g., spam detection)
Understanding these fundamental concepts is essential as you explore more advanced supervised learning techniques in the following sections.

​
What’s Next

Now that you understand supervised learning fundamentals, you’re ready to dive deeper into specific algorithms and techniques:
  • Learn about regression models and how to fit them to data
  • Understand gradient descent for optimizing model parameters
  • Explore multiple linear regression with many features
  • Study logistic regression for classification problems
  • Address overfitting to improve model performance

Build docs developers (and LLMs) love

Get started for freeTalk to us