Chapter 1: Classical ML
Gaussian mixtures, EM and the silhouette
You will split a point between two groups with EM, see k-means as a special case, and score a split with the silhouette. Interviewers ask about all three.
What you will learn
- Explain EM as soft k-means
- Score clusters with the silhouette and say what it cannot prove
Key terms
- Gaussian mixture
- a model that draws each group as a bell curve
- responsibility
- the fraction of a point that belongs to one bell
- silhouette
- a score from -1 to 1: own group against the nearest other group
Prerequisites
Take Clustering: k-means, restarts and k-means++ first. It comes right before this lesson in the course.
Where this lesson picks up: Last lesson, k-means gave each point wholly to its nearest centre. What if the groups overlap?
What comes next
Clusters group the rows. Next: can we also shrink the columns?
Next lesson: PCA: keeping the most spread
More lessons in this part
- Clustering: k-means, restarts and k-means++ 15 min
- PCA: keeping the most spread 7 min
- Capstone: churn with survival analysis 16 min
Take the lesson
Short cards, one idea at a time, with quick checks as you go.