Mises Dev on Nostr: I'm a few steps behind, my project is still a bit primitive haha Right now, it only ...
I'm a few steps behind, my project is still a bit primitive haha
Right now, it only performs a simple breadth-first search (BFS) on the graph, where I calculate the similarity between user names and the search term using an algorithm I implemented based on this paper: http://www.catalysoft.com/articles/StrikeAMatch.html
I'm not using any database; I implemented everything from scratch in C to achieve the highest possible performance. By using the paper above and implementing it in the most optimized way possible with bitwise operations and everything, the similarity calculation became orders of magnitude faster than all existing methods—about 300% faster than cosine similarity, for example.
But it's still a pretty primitive project, as I mentioned. Right now, it's just a simple search in a graph and list. Later on, depending on the needs that come up, I might implement something like PageRank when considering user posts. For now, the search is just done by profile.
Right now, it only performs a simple breadth-first search (BFS) on the graph, where I calculate the similarity between user names and the search term using an algorithm I implemented based on this paper: http://www.catalysoft.com/articles/StrikeAMatch.html
I'm not using any database; I implemented everything from scratch in C to achieve the highest possible performance. By using the paper above and implementing it in the most optimized way possible with bitwise operations and everything, the similarity calculation became orders of magnitude faster than all existing methods—about 300% faster than cosine similarity, for example.
But it's still a pretty primitive project, as I mentioned. Right now, it's just a simple search in a graph and list. Later on, depending on the needs that come up, I might implement something like PageRank when considering user posts. For now, the search is just done by profile.