- Learn
- Machine Learning
- Unsupervised Learning
ML-M07
Unsupervised Learning
Finding structure without labels: clustering, dimensionality reduction, visualization of embeddings, anomaly detection and association rules.
Most data in the world has no labels. Nobody has tagged your customers by type, your sensor readings as normal or broken, or your documents by topic. Unsupervised learning finds the structure that is already there: groups, directions of variation, rare oddities and things that occur together.
You will learn k-means, hierarchical clustering and DBSCAN, Gaussian mixtures fitted with expectation-maximization, principal component analysis, t-SNE and UMAP for visualizing high-dimensional data, the main anomaly detection methods, and association rules for market basket analysis.
Lessons
- ML-07.01 k-means clustering k-means groups points into k clusters by alternating two steps: assign each point to its nearest centroid, then move each centroid to the mean of its points, until nothing changes. Beginner 14 min
- ML-07.02 Hierarchical clustering and DBSCAN Hierarchical clustering builds a tree of merges, shown as a dendrogram, and you cut it at the level that gives the clusters you want. Intermediate 12 min
- ML-07.03 Gaussian mixture models and EM A Gaussian mixture model assumes data comes from several overlapping bell-shaped clusters and gives each point a probability of belonging to each one, a soft clustering. Advanced 12 min
- ML-07.04 Principal Component Analysis PCA finds the directions in which data varies most and projects data onto the top few of them, reducing dimensions while keeping as much variance as possible. Intermediate 14 min
- ML-07.05 t-SNE and UMAP t-SNE and UMAP map high-dimensional data into two or three dimensions for visualization, keeping similar points close together. Intermediate 10 min
- ML-07.06 Anomaly detection Anomaly detection finds rare, unusual data points, such as fraud or failing machines, often without labels. Intermediate 12 min
- ML-07.07 Association rules and market basket analysis Association rule mining finds items that frequently occur together, such as bread and butter in shopping baskets. Beginner 8 min