📄️ Unsupervised learning
Discovering structure in unlabeled data: clustering, dimensionality reduction and anomaly detection. 6 hours, applied projects, a 40-question exam and a verifiable certificate.
📄️ 1. Searching without labels
Module 1 of the Unsupervised Learning premium course: what the objective becomes when the target disappears, the three families of tasks, and the problem of evaluating without ground truth.
📄️ 2. k-means
Module 2 of the Unsupervised Learning premium course: the k-means algorithm step by step, the inertia it minimizes, the role of k-means++ and the cluster shapes it cannot see.
📄️ 3. Choosing the number of clusters
Module 3 of the Unsupervised Learning premium course: why inertia is not enough, how to read the elbow method, what the silhouette score measures and how to arbitrate with the business.
📄️ 4. Hierarchical clustering
Module 4 of the Unsupervised Learning premium course: the agglomerative principle, linkage criteria and their effects, and reading a dendrogram to choose where to cut.
📄️ 5. DBSCAN
Module 5 of the Unsupervised Learning premium course: the notion of density, the eps and min_samples parameters, core, border and noise points, and tuning via the k-distance graph.
📄️ 6. Principal component analysis
Module 6 of the Unsupervised Learning premium course: what PCA does, reading explained variance, choosing the number of components and interpreting the axes.
📄️ 7. t-SNE and UMAP
Module 7 of the Unsupervised Learning premium course: what t-SNE and UMAP are for, the role of perplexity and neighbors, and above all what a projection never lets you conclude.
📄️ 8. Anomaly detection
Module 8 of the Unsupervised Learning premium course: defining an anomaly, univariate statistical approaches, the isolation forest, and evaluation when anomalies are rare.
📄️ 9. Gaussian mixture models
Module 9 of the Unsupervised Learning premium course: probabilistic assignment, flexible cluster shapes, the EM algorithm and choosing the number of components with BIC.
📄️ 10. Segmentation project
Module 10 of the Unsupervised Learning premium course: the complete workflow of a segmentation, from framing and variable choice to comparing methods, characterizing and documenting the segments.
📄️ Recap and exam
Complete recap of the Unsupervised Learning premium course: clustering, dimensionality reduction and anomaly detection module by module, then the 40-question exam.