Initializations of Kmeans Suppose a data sets is generated by sampling examples uniformly at random from r spherical gaussians with an std of 1. In which cases is Kmeans with Kmeans++ initialization likely to be significantly better than Kmeans with standard initialization?

Options (can be multiple): 1. The clusters are very close each other. 2. The clusters are far from each other. 3. r is large 4. r is small 5. All clusters have equal probability 6. One cluster has much higher probability than the others.

j The clusters are very close each other. :] The clusters are far from each other. :] ris large 3 r is small 3 All clusters have equal probability 3 One cluster has much higher probability than the others. $ SmeiT You have used 0 of 5 attempts S