May 2018
Beginner
490 pages
13h 16m
English
There are several variations of Lloyd's algorithm. But all of them follow a common philosophy.
For a given xn (data point), the distance from the centroid µ in its cluster must be less than going to another center. Exactly like how a person in the lunch example wants to be closer to one table rather than having to go far to get a sandwich because of the crowd!
The best centroid µ for a given xn is as follows:

This calculation is done for all µ (centroids) in all the clusters from k1 to K.
Once each xi has been allocated to a Kk, the algorithm recomputes µ by calculating the means of all the points that belong to each cluster ...
Read now
Unlock full access