CS 461: Machine Learning
Instructor: Kiri Wagstaff

Reading Questions for Lecture 8

Clustering (Ch. 7.1-7.4, 7.8)
  1. (Gavin) From page 144: k-means clustering is a special case of EM. In geometric terms, k-means can be viewed as a circle, and EM as an ellipse. Can k-means's single parameter be considered the radius of the circle, and EM's two parameters the defining values (whatever they're called) for the ellipse?
  2. (Matthew) When is clustering most effective at preprocessing data?
  3. (Ron) Explain a little better that an increase in expectation implies an increase in the incomplete likelihood?
  4. (Roice) From the extra credit problem, what is the COBWEB clustering algorithm?