You are viewing a single comment's thread from:
RE: How-to solve SPAM and Democratize Steem: Introducing UserAuthority
While I'm largely undecided on the details of how this could be implemented, if it should, and specifically into what) - I think there's a discussion worth having here.
Using the quality of a user (based off followers) to gauge their rewards/impact is a relatively new concept that I'm seeing used more and more in some of the 3rd party services. Based on how they're performing - it might be good to apply the concept to the overall Steem ecosystem or perhaps SMTs.
spot on mate.. speaking volume here.. upVoted so as not missed! @cnts :]
Hi Jesta, via DMs, 2 people suggested to me that I should forward this via email to somebody called "sneak". Do you know if that's the same person as account @sneak ? Do you know this person "sneak", and could you perhaps ask him/her to have a look? Or do you think I should send the email?
Yes, that'll be the one.
Yup, @sneak is one of the guys on the Steem team.
Have you considered applying the SVD calculation to speed up computation and basing your calculation on the top 90% of the eigenvalues?
No, not yet and I don't think it would matter that much. As comparing users to pages, on steemit, only a very small amount of users (let's say 500k) compared to pages on the web (a lot more) exist. Computationally it doesn't currently seem to be a problem, so optimizing it would bring little added value. And also, the total follower graph doesn't change drastically daily: a lot of new relationships may be formed and some (a lot less) are terminated, yet all in all that wouldn't make a big difference to the overall end-results. So daily re-calculating the UA binary index (containing about 8 bytes per user, or more if more information is stored in it), is fine.