k-means++

noun

say it: “KAY MEENZ PLUS PLUS” /ˈkeɪ ˈminz ˈplʌs ˈplʌs/

A method for choosing the starting centers in k-means so they are spread out, which usually leads to better groups.

What is k-means++?

k-means++ chooses k-means' starting centers so they're spread out: the first at random, then each next one more likely to be far from those already chosen. Spread-out starts make a poor local minimum less likely and need fewer rounds. It's scikit-learn's default (init="k-means++").

In a sentence

  • “Because k-means++ is the default, she never had to set init herself.”
  • “k-means++ put the starting centers far apart, so no two began in the same group.”

Level: Freelancer · More AI Basics terms at this level

Updated