Clustering, dimensionality reduction, density estimation and anomaly detection without labels to check yourself against.
8h · 7 lessons · 1 challenge
Before this: Linear Algebra for Machine Learning, Regression & the Supervised Setup
Without labels there is no accuracy to hide behind, so evaluation becomes the hard part. This module covers k-means, mixture models, PCA, t-SNE and UMAP, and is unusually strict about how each one is validated and how each one misleads.
Anomaly detection closes it out, framed as the production problem it usually is.
Two-dimensional projections of known structures. Say what the real structure was — and learn how much t-SNE invented.
Worth 350 XP.
