Partitioning Around Medoids (PAM)

Implementation of Partitioning Around Medoids (PAM) in H2O-3

Pros:

• Deterministic in nature
• Outperforms k-means for some applications

Cons:

• Inherently O(n^2) time and possibly space depending on implementation

References:

• pp.102-104 of Finding Groups in Data by Kaufman and Rousseeuw.
• https://en.wikipedia.org/wiki/K-medoids
• https://en.wikibooks.org/wiki/Data_Mining_Algorithms_In_R/Clustering/Partitioning_Around_Medoids_(PAM)