Skip to main content
. 2017 May 15;375(2096):20160293. doi: 10.1098/rsta.2016.0293

Figure 1.

Figure 1.

K-means clustering based on a Gaussian-like data distribution assumption is largely unsuccessful for other shapes of data clouds. Only if the two objects move further away from each other can they be considered as Gaussian clouds and be properly separated. Dashed line denotes separating the computed cluster boundaries; filled dots, cluster centres.