Machine learning includes several major learning paradigms, each suited to different kinds of problems and available data. Supervised learning uses labeled examples containing inputs and target outputs. It commonly supports classification and regression. Unsupervised learning works with data without externally provided target labels and attempts to discover structures such as clusters, associations, or lower-dimensional representations. Semi-supervised learning combines a relatively small labeled dataset with a larger unlabeled dataset. Self-supervised learning creates learning targets from the data itself, often by predicting hidden or transformed parts of an input. Reinforcement learning involves an agent interacting with an environment, receiving rewards or penalties, and learning a policy that aims to maximize cumulative reward. These categories describe learning setups rather than mutually exclusive families of algorithms. For example, neural networks can be used in supervised, self-supervised, and reinforcement learning. The choice of paradigm depends on the problem, the type and quantity of available data, the feedback signal, and the desired behavior. Understanding these differences helps practitioners choose appropriate methods and evaluation strategies.
Types of Machine Learning
Machine learning is commonly divided into supervised, unsupervised, semi-supervised, self-supervised, and reinforcement learning according to the type of data, feedback, and learning signal used.
Explanation
Variables
| Symbol | Meaning |
|---|---|
Supervised Learning | Learning from labeled input-output examples. |
Unsupervised Learning | Learning patterns or structures from data without externally supplied target labels. |
Semi-Supervised Learning | Learning using a combination of labeled and unlabeled examples. |
Self-Supervised Learning | Learning from targets generated automatically from the input data. |
Reinforcement Learning | Learning behavior through interaction with an environment and reward feedback. |
Example or application
A company may use supervised learning to predict customer purchases from labeled historical records, unsupervised learning to discover customer groups, semi-supervised learning when only some records are labeled, self-supervised learning to pretrain a language model, and reinforcement learning to optimize sequential decisions.
Text
Different machine learning approaches use different forms of information to learn. Some rely on labeled examples, others discover patterns without labels, and reinforcement learning uses feedback from interactions with an environment.
Key Points
Supervised learning uses labeled examples.
Unsupervised learning discovers patterns without externally supplied target labels.
Semi-supervised learning combines labeled and unlabeled data.
Self-supervised learning generates training targets from the data itself.
Reinforcement learning uses interaction and reward feedback.
The same model architecture may be used with different learning paradigms.
Applications
Predictive modeling
Customer segmentation
Image and language pretraining
Autonomous decision-making
Recommendation systems
Robotics and control
Learning Outcomes
Identify the five major machine learning paradigms.
Compare the type of training signal used by each paradigm.
Select an appropriate learning paradigm for a simple problem.
Understand that learning paradigms are not the same as model architectures.
Practice Questions
What are the five major types of machine learning?
How does supervised learning differ from unsupervised learning?
What is the difference between semi-supervised and self-supervised learning?
How does reinforcement learning receive feedback?
Related Concepts
Supervised Learning
Unsupervised Learning
Semi-Supervised Learning
Self-Supervised Learning
Reinforcement Learning
Common Mistakes
Confusing semi-supervised learning with self-supervised learning.
Assuming unsupervised learning never uses any form of objective or feedback.
Treating reinforcement learning as ordinary supervised learning with reward labels.
Assuming each learning paradigm requires a completely different neural network architecture.
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Goodfellow, I., Bengio, Y., and Courville, A. Deep Learning.
- Bishop, C. M. Pattern Recognition and Machine Learning.
- Sutton, R. S. and Barto, A. G. Reinforcement Learning: An Introduction.
Suggest a correction or source · More in Artificial Intelligence & Machine Learning