Types of Machine Learning Algorithms

1. Introduction

Learning algorithms differ in the feedback they receive and the goal they optimize. Choosing a family begins with the task and available labels, not with the popularity of an algorithm.

2. Core concepts

Supervised learning uses labeled examples for regression or classification. Unsupervised methods search for structure without a target, such as clusters or lower-dimensional representations. Reinforcement learning learns actions from rewards across interactions and is not simply a classifier with a different name.

3. Topics covered in this module

• Machine-learning history and the role of data and compute
• Supervised regression, classification, and label requirements
• Unsupervised clustering, dimensionality reduction, and anomaly detection
• Reinforcement learning, state, action, reward, and exploration

4. Syntax and how to read it

task = 'predict a numeric delivery duration'
learning_type = 'supervised regression'
metric = 'mean_absolute_error'

This task definition connects the target type to a model family and interpretable metric before selecting an estimator.

5. Practical workflow

Specify the output needed, whether trustworthy labels exist, and how success will be measured. Start with the least complex method that answers the question and can be evaluated responsibly.

6. Common mistakes

• Calling clusters ground truth categories without validation.
• Using reinforcement learning when a static supervised dataset and decision rule are enough.

7. Best practices

• Choose algorithms from target type, available feedback, constraints, and evaluation.
• Explain the assumptions and limitations of an unsupervised grouping.

8. Practice task

You have labeled examples of house prices and want to estimate a price for a new house. Which broad learning type applies?

9. One possible solution

Supervised regression, because the training examples include a numeric target value.

10. Summary

Compare supervised, unsupervised, and reinforcement learning and select a method from the task. Choose algorithms from target type, available feedback, constraints, and evaluation. Explain the assumptions and limitations of an unsupervised grouping.

11. What comes next

The next module, Practical ML using Scikit-learn, builds on this foundation and applies it to a new part of the analytics workflow.