Local Minimum

noun

say it: “LOH-kul MIN-ih-mum” /ˈloʊkəl ˈmɪnəməm/

A point where a value is lower than at all nearby points, but not the lowest possible overall.

What is a local minimum?

A local minimum is a result that small changes can't improve but that isn't the best possible, like the bottom of a small dip when a deeper valley lies elsewhere. k-means can settle in one, depending on where its centers start, so scikit-learn runs it from several starts (n_init) and keeps the best.

In a sentence

  • “One start left k-means in a local minimum, with two centers sharing a single group.”
  • “Running ten starts made a poor local minimum less likely.”

Level: Freelancer · More AI Basics terms at this level

Updated