K-means

Question 1

What type of learning method is k-means?

Answer 1

Unsupervised learning

Question 2

What does an unsupervised learning method mean?

Answer 2

> All the data is unlabeled
This means that there is no correct answer
We may not know the answers

Question 3

What is clustering?

Answer 3

> This is dividing a data set into groups which are similar to each other
There are lots of different ways of dividing a data set
We can plot graphs of the results

Question 4

Why is scale imporant for k-means?

Answer 4

Because clusters may have sub-clusers if you look closer

Question 5

What is the k-means method?

Answer 5

1. Chose the number of clusters (k)
2. Place k-centroids randomly
3. Identify the points closest to the centroid
4. Compute the new centroid clusers
5. Repeat until centroids do not move

Question 6

What are the pros of k-means?

Answer 6

> Very easy to implement

> Easy to explain

Question 7

What are the cons of k-means?

Answer 7

> You have to chose the number of clusters

> Subject to local minima because of the initial positions of the centroids (so random restarts are required)