Unsupervised learning uses unlabeled data to discover patterns, structures, or groupings without explicit outcome labels. The model “learns” by identifying similarities, clusters, or latent structures within the data, somewhat analogous to how humans can notice patterns without being told the correct answer. In AAIA’s fundamentals coverage, unsupervised methods (e.g., clustering, dimensionality reduction) are explicitly linked to situations where labels are unavailable or costly , yet insight is still needed.

Supervised learning (A) requires labeled examples (input–output pairs). Federated learning (B) describes a distributed training paradigm, not the labeling requirement itself. Reinforcement learning (C) uses feedback in the form of rewards and penalties rather than unlabeled static datasets. Therefore, unsupervised learning is the correct type that directly uses unlabeled data to learn structure.