You are viewing a single comment's thread from:
RE: Improving Steem’s rankings to cater to diverse content preferences
I would not even try to map the data to the Grassmannian and use the naturally defined metric to determine clustering of the data on various manifolds.
Bad ideas don't work in practice.
Seriously. Avoid this notion at all costs. Grassmannians are tricky beasts.
Is it the intermediate mean for Dₛₜ? You think I should cluster directly on the votes? That worry crossed my mind but I hadn't yet had time to delve into the ramifications. I am not familiar with why they are tricky.
I dunno man. Anytime I start thinking about the space of k-dimensional linear subspaces on an n-dimensional space (i.e. Gr(n,k)), and about points on that space, my mind begins to get warped.
The big question is how can other metrics be used as a way of identifying clustered points on a manifold.
I think we'll see how persistent homology will be coming into play.