Back to GlossaryCore Concepts

Unsupervised Learning

Definition

A machine learning approach where models discover patterns and structure in data without labeled examples or predefined outputs.

Unsupervised learning finds hidden patterns in data without the guidance of labeled examples. Common techniques include clustering (grouping similar data points, e.g., K-means), dimensionality reduction (compressing data while preserving structure, e.g., PCA), and density estimation. Autoencoders and variational autoencoders are neural network approaches to unsupervised learning. Self-supervised learning, a modern variant used in pre-training large language models, creates its own labels from the data itself (e.g., predicting masked words). Unsupervised learning is valuable because labeled data is expensive to create, while unlabeled data is abundant.

Companies in Core Concepts

View Core Concepts companies →