It was observed that after an initial phase of rapid movement of the center points, the k-means algorithm tends to settle into a long phase where the center points move very slowly. This suggests that a smart algorithm should attempt to update nearest neighbors incrementally after the center moves, rather than recompute them from scratch each time. An algorithm has been especially developed for this purpose, and is analyzed as a function of the distances that the centers move.
|State||Published - 1999|
|Event||Proceedings of the 1999 10th Annual ACM-SIAM Symposium on Discrete Algorithms - Baltimore, MD, USA|
Duration: 17 Jan 1999 → 19 Jan 1999
|Conference||Proceedings of the 1999 10th Annual ACM-SIAM Symposium on Discrete Algorithms|
|City||Baltimore, MD, USA|
|Period||17/01/99 → 19/01/99|