Re: Question 14 LLoyd's algorithm  empty clusters
I'm starting Lloyd's algorithm over anytime it comes up with fewer clusters than desired, so that each iteration I'm using the full number. Thanks to this original post I realized that this is an issue. Usually Lloyd's algorithm comes up with the correct number of clusters, but occasionally something weird happens. The larger the value of K (# desired clusters), the more frequently one ends up with less than the requisite number. I was very surprised to see rare cases where only one cluster remained (Perhaps I SHOULD be surprised, as all I've noticed is a bug in my code!). Anyone else see Lloyd's algorithm occasionally (rarely) coming up with only one region?
Thanks Mark, for asking the question. Also thanks to our illustrious professor!
:)
