Unsupervised learning is a machine learning approach in which algorithms analyze data without receiving externally supplied target labels for each training example. Instead of learning to predict a known answer, the algorithm attempts to discover useful structures within the data. Clustering is a common unsupervised task that groups similar observations together. Algorithms such as K-means and hierarchical clustering can identify groups based on similarity measures. Dimensionality reduction methods, including Principal Component Analysis (PCA), transform data into a smaller number of features while attempting to preserve important information. Other unsupervised tasks include association-rule discovery, density estimation, and anomaly detection. The results of unsupervised learning may require interpretation because the discovered groups or patterns do not automatically correspond to meaningful real-world categories. Unlike supervised learning, the training dataset does not normally provide a target answer for every example. However, unsupervised algorithms still use mathematical objectives, similarity measures, or structural assumptions to guide learning. Their usefulness depends on data quality, feature representation, algorithm selection, and the evaluation method.
Unsupervised Learning
Unsupervised learning is a machine learning paradigm that discovers patterns, structures, or relationships in data without relying on externally provided target labels.
Clustering Objective: Minimize Σᵢ Σₓ∈Cᵢ d(x, μᵢ)² (K-means objective)
Explanation
Variables
| Symbol | Meaning |
|---|---|
Cᵢ | The i-th cluster containing a group of observations. |
μᵢ | The centroid or mean of cluster Cᵢ in K-means. |
d(x, μᵢ) | The distance between observation x and cluster centroid μᵢ. |
k | The number of clusters specified in K-means. |
Example or application
A retail company has customer purchase records without predefined customer categories. An unsupervised clustering algorithm can group customers with similar purchasing behavior, helping the company identify different customer segments.
Text
Unsupervised learning explores unlabeled data to identify natural groupings, relationships, unusual observations, or simpler representations of complex datasets.
Key Points
Unsupervised learning generally uses data without externally supplied target labels.
Clustering groups observations based on similarity.
Dimensionality reduction creates simpler representations of data.
K-means is a common clustering algorithm.
Discovered patterns may require human interpretation.
An unlabeled dataset can still be used with a defined mathematical learning objective.
Applications
Customer segmentation
Anomaly detection
Document grouping
Exploratory data analysis
Image compression and dimensionality reduction
Market basket analysis
Learning Outcomes
Define unsupervised learning.
Explain how clustering discovers groups in data.
Describe dimensionality reduction.
Identify common unsupervised learning algorithms.
Explain why discovered clusters may need interpretation.
Practice Questions
What is unsupervised learning?
How does clustering work?
What is the purpose of dimensionality reduction?
Why might an unsupervised learning result require human interpretation?
Related Concepts
Types of Machine Learning
Clustering
K-means
Principal Component Analysis
Anomaly Detection
Common Mistakes
Assuming unsupervised learning means that no objective function is used.
Treating every discovered cluster as a meaningful real-world category.
Confusing clustering with classification.
Assuming K-means automatically determines the correct number of clusters.
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.
- Bishop, C. M. Pattern Recognition and Machine Learning.
- Hastie, T., Tibshirani, R., and Friedman, J. The Elements of Statistical Learning.
- Murphy, K. P. Probabilistic Machine Learning: An Introduction.
Suggest a correction or source · More in Artificial Intelligence & Machine Learning