Weights in dbscan()
EmmanuelGustin opened this issue · comments
Hi,
Changing back from my comment earlier. Some of the unexpected behavior that I saw was due to bugs in my own code -- a good thing!
Still one thing remains that I find odd: Using uniform random weights does improve clustering... at least in my data set.
But the issue can be closed!
I am glad to hear that it is not a problem on our end!
About the random weights: What you do is basically "soft" sampling, meaning that you randomly remove weight from some data points. This effectively has a similar effect as increasing minPts and maybe that is what drives what you see.