mengxr··on Introducing Streaming K-Means in Spark MLlib 1.2Now we only detect dying clusters while keeping k constant. Dynamic values of k would be a nice feature to add in later releases:)