dm.cs.tu-dortmund.de/en/mlbits/cluster-kmeans-extensions/
Extensions of k-means Clustering – Lecture Notes
rt ^2\) .
That does not imply the mean minimizes other distances!
Counter example : \(1,2,3,4,10\in \mathbb {R}^1\) ; Euclidean distance becomes \(\smash {\sqrt{\smash {(x_i-m)^2}\mathstrut }}=|x_i-m|\) [...] t x_i\right\rVert =1\)
at each iteration, the new centers are normalized to \(\mu _C \leftarrow \mu _C/\left\lVert \mu _C\right\rVert\) such that \(\left\lVert \mu _C\right\rVert =1\)
normalized \(\mu [...] dimension [ BrMaSt96 ] for use with Manhattan norm \(\left\lVert \smash {x_i-m_j}\mathstrut \right\rVert _1\)
\(k\) -midranges : use \(\tfrac 12(\max _i - \min _i)\) as center, for use with maximum norm [ CaCh98 …
