Does K-means clustering algorithm really finds the global minimum or not?

machine_learning
k-meansclustering
data_science

#1

Hello,

The K-means clustering algorithms uses the square of the Euclidean distance
to find the global minimum solution and this problem is not trivial. Does this mean that the K-means algorithm only tries to find the global minimum solution and it is not guaranteed that it will actually find it?
If its true, how can we rely on the accuracy of the algorithm?

Thanks


#2

Hi Aditya,

It’s correct that it only tries to find a optimum solution, it could be a local or global maxima. To make sure you reach to the global maxima is to repeat the clustering exercise multiple times and take the most optimum solution.

Regards,
Aayush