1 · Dimensionality reduction
From the curse of dimensionality to PCA — compress features without losing the signal.
●Track 6 · live
Unsupervised ML
Find structure without labels — clustering and projection.
PCA, K-Means, hierarchical clustering, DBSCAN, silhouette, anomaly detection (Isolation Forest, LOF). The K-Means iteration scene shows centers and assignments converging in real time.
9 / 15 lessons3 chapters
2 · Clustering
K-Means as the canonical clusterer, then hierarchical and DBSCAN for the shapes K-Means can't handle.
3 · Anomaly detection & evaluation
Isolation Forest and LOF for finding outliers — plus the metrics that tell you whether your unsupervised model is working.