Unsupervis

Notes

Unsupervised Learning, Clustering, & Dimensionality Reduction Nov. 18, 2025 - Lecture 9 DS3000 - Intro to Machine Learning Outline • What is unsupervised learning? • Clustering • Clustering algorithms: K-means, GMM, EM algorithm, Agglomerative Clustering, DBSCAN • What is dimensionality reduction? Why do we need it? • PCA, UMAP, t-SNE, Autoencoders •Supervised vs. Unsupervised Learning* Supervised learning uses data with labels, while unsupervised learning involves only data without labels, aiming to uncover hidden patterns or structures. The main tasks in unsupervised learning are clustering, association rules, and dimensionality reduction. •Clustering* Clustering groups unlabeled data into clusters based on similarities or differences. For example, in the figure on the left, three clusters are visible. In practice, defining a similarity measure is essential for grouping points effectively. •Clustering Applications* Clustering is useful across industries: - Market segmentation - Social network analysis - Medical imaging - Image segmentation - Anomaly detection •Types of Clustering* Defining a "cluster" is challenging, leading to various...