Unsupervised learning

Unsupervised learning is a type of machine learning where algorithm are trained using unlabeled data. Key characteristics:

• Does not require output categories or targets to be predefined.
• Derives structure and relationships directly from the data.
• Aims to model the underlying probability density of the inputs.
• Useful for exploratory analysis and pattern discovery.

Major categories of unsupervised learning techniques:

• Clustering algorithms group data points based on similarity.
• Dimensionality reduction techniques compress data while preserving structure.
• Density estimation methods estimate data distribution.
• Anomaly detection identifies outliers that differ from expected patterns.

Applications include market segmentation, social network analysis, astronomical data analysis, and more.

Unlike supervised learning, unsupervised techniques must infer patterns solely from datasets without explicit feedback. Performance is evaluated based on utility for exploration and analysis rather than predictive accuracy.

See also:

• Machine learning
• Supervised learning
• Cluster analysis
• Principal component analysis