Explain why K$-$mean may not find the global optimum $($minimum of within cluster variance: sum of the squared errors$),$ but find local minimums when initialization differently $(10$ points$).($Try different initializations for the data above, get the cluster results, and compute sum of the squared errors$)$