Nostr Web Client

First off, GrapeRank (GP) should only be compared to Personalized Pagerank (PP). Both are personalized, or "local".

Second, I very much disagree that PP is about popularity. It depends on the graph it is run. If you would "run it on zaps", you would get something that is closer to "value" than "popularity".

Third, the mathematical properties of GP are unknown (does it converge? what's the relationship with distance? what's the relationship with flux? is it graph stable?), but I would they are not promising. Here for example:

Npubs that are 2 and 3 hops from the user, span the whole spectrum of possible values. This suggests that distance DOES NOT really play a role in GP. But distance is something the user can influence directly, by following/unfollowing!

Compare it to Personalized Pagerank where distance actually has both a theoretical and measurable impact. As you can see, the tendency is that, the furthest you go, the lower the score.

david 11mo ago 💬 1

The average GR score decreases as hops increases, as it does with PageRank. Distance plays a role with both.

Reply to this note

Please Login to reply.

Discussion

Pip the WoT guy 11mo ago 💬 1

The average does, but the variance is really high

david 11mo ago

A high variance does not imply GR is not useful. I’d argue the opposite: it implies the GR score conveys information that is not conveyed by the DoS score alone.

A low variance (tight correlation) between the PR score and the DoS score would imply we may as well just scrap PR and use the DoS score which is easier to calculate. Although I don’t think the correlation is tight enough to suggest we should do that.