C5Unsupervised learning

Unsupervised learning looks for structure when there are no labels to lean on: groups of similar points, or a low-dimensional summary of high-dimensional data. It is how you explore a fresh dataset, compress it, and build intuition before any predictive model exists, and it is often the first thing you reach for when you do not yet know what question to ask.

You will cluster with k-means (seen as alternating minimization), improve it with k-means++ and principled choices of k, and handle non-Euclidean or noisy data with k-medoids, DBSCAN, and HDBSCAN. You will read dendrograms from hierarchical clustering, fit Gaussian mixture models with the EM algorithm (soft k-means with covariances), and visualize high-dimensional structure with t-SNE and UMAP while learning where they mislead.

These tools power exploratory analysis, customer and image segmentation, and feature compression across the field. The EM algorithm reappears throughout latent-variable modeling, and the clustering and quantization ideas here are exactly what large-scale vector search builds on later.