Open michaelaye opened 8 years ago
We can't just use total noise compared to total members of clusters, because then 1 bad user could skew this a lot. We need to scale this somehow by number of users creating that noise.
We can't just use total noise compared to total members of clusters, because then 1 bad user could skew this a lot. We need to scale this somehow by number of users creating that noise.