Chapter 1: Classical ML
Clustering: k-means, restarts and k-means++
You will run k-means by hand, see why it can stop stuck, and pick better starts with k-means++. Interviewers ask about all three.
What you will learn
- Run Lloyd's algorithm and explain why restarts matter
- Explain k-means++ seeding
Key terms
- centroid
- the mean of the points in one group
- WCSS
- total squared distance from each point to its group's centroid
- Lloyd's algorithm
- repeat: each point joins its nearest centre, then each centre moves to its group's mean
- k-means++
- a way to pick starting centres that sit far apart
Prerequisites
Take Explaining a model: permutation importance and SHAP first. It comes right before this lesson in the course.
Where this lesson picks up: So far every row had a label to predict. Now the rows have no labels. How can you still sort them into groups?
What comes next
k-means gives each point to exactly one group. Next: what if the groups overlap?
Next lesson: Gaussian mixtures, EM and the silhouette
More lessons in this part
- Gaussian mixtures, EM and the silhouette 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.