pippellia on Nostr: I am using a different idea with PP. First off, it's a monte-carlo version, so I am ...
I am using a different idea with PP. First off, it's a monte-carlo version, so I am simulating a bunch of random walks.
Secondly, I've added a new stopping rule: break the walk when a cycle is found.
example: a walk is a bunch of IDs of nodes, like {1, 11, 234, 7, 0}
The walk {1,2,1,2,1,2} contains cycles, so it's stopped earlier at {1,2}. This means 1 and 2 gets visited less often.
Cycles are really improbable for normal users, but highly probable for spam networks. The net result that I've found is basically no change for normal people, and a big de-rank for such networks (their scores go down by a factor of ~3)
Secondly, I've added a new stopping rule: break the walk when a cycle is found.
example: a walk is a bunch of IDs of nodes, like {1, 11, 234, 7, 0}
The walk {1,2,1,2,1,2} contains cycles, so it's stopped earlier at {1,2}. This means 1 and 2 gets visited less often.
Cycles are really improbable for normal users, but highly probable for spam networks. The net result that I've found is basically no change for normal people, and a big de-rank for such networks (their scores go down by a factor of ~3)