12. Clustering by MIT OpenCourseWare

Summary by www.lecturesummary.com: 12. Clustering by MIT OpenCourseWare

Clustering in Machine Learning
This content, presumably from an MIT OpenCourseware lecture on machine learning, concentrates on clustering as an unsupervised learning method. It defines the fundamental concept of clustering as an optimization problem aiming to minimize dissimilarity between clusters subject to a constraint such as a fixed number of clusters (K).
Clustering Algorithms
The lecture compares two prominent clustering algorithms:

• Hierarchical Clustering: A deterministic but possibly slow algorithm generating a dendrogram showing cluster mergers.

• K-Means: A slower but non-deterministic algorithm needing pre-choice of K and being vulnerable to initial centroids.

Scaling Attributes
The application of scaling attributes, such as z-scaling, is covered to solve problems with features' varying ranges affecting distance calculations. An example is provided on a medical patient dataset.