Hi @brugeman — I’m looking at trust.nostr.band and curious to know where you’d like to take it. I think using PageRank or something like it as the foundation for nostr search is a great idea.

Is there a reason you use the iterative method rather than the power method (use matrix algebra to solve for the principal eigenvector)? My understanding is that the power method is computationally more efficient than iterative. 

 The power method is iterative (like many other methods). I was myself confused by his use of the word "iterative" . 

 I was looking at this when I wrote the above 

https://en.wikipedia.org/wiki/PageRank#Power_method 

 mmm. They are the same method, but the second is more efficient in practice because the libraries for matrix multiplication are highly optimized in basically all languages.

However, one should be careful in using efficient sparse matrices representations (like CSR), otherwise the power method might use more memory. 

 I'm probably just too bad at math. Also I'm not sure how well it would scale to 100M nodes.