Timeline for Termination conditions for K-means and their interconnection
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 4 at 4:39 | history | edited | wjktrs | CC BY-SA 4.0 |
added 3 characters in body
|
| Feb 17 at 20:37 | history | edited | wjktrs | CC BY-SA 4.0 |
deleted 15 characters in body
|
| Feb 2 at 21:52 | history | edited | wjktrs | CC BY-SA 4.0 |
added 4 characters in body
|
| Feb 2 at 14:55 | comment | added | wjktrs | See modified answer - I really meant that object swapping ("individual assignment") stops, not change in cluster sample size. Also, there are numerous variants of K-means, as well as fuzzy K-means. | |
| Feb 2 at 14:54 | history | edited | wjktrs | CC BY-SA 4.0 |
deleted 88 characters in body
|
| Feb 2 at 10:56 | comment | added | Christian Hennig | "K-means stops when there is no change in the number of objects assigned to clusters -- i.e., change in cluster sample size." This is very strange as assignment changes in individual points can in principle be quite substantial in a situation in which cluster sizes don't change. The algorithms I have seen stop when no individual assignment changes anymore. You may refer to a specific algorithm or implementation. I'd be surprised if the classic algorithms (Lloyd's etc.) are presented anywhere as stopping based on cluster sizes. | |
| Jan 22 at 5:36 | history | edited | wjktrs | CC BY-SA 4.0 |
added 559 characters in body
|
| Jan 22 at 5:30 | history | edited | wjktrs | CC BY-SA 4.0 |
added 559 characters in body
|
| Jan 22 at 0:02 | history | edited | wjktrs | CC BY-SA 4.0 |
added 2 characters in body
|
| Jan 21 at 23:57 | history | edited | wjktrs | CC BY-SA 4.0 |
added 1050 characters in body
|
| Jan 21 at 19:49 | history | answered | wjktrs | CC BY-SA 4.0 |