What’s the typical vector dimension/size that this would work well for? I did something naively like this using array indexes of about length 20 vectors on Clo

What’s the typical vector dimension/size that this would work well for? I did something naively like this using array indexes of about length 20 vectors on Cloudant (CouchDB) in order to implement a text similarity search, based on something simple like LDA auto tagging and keyword frequency clustering vectors. I would query based on input text vectorized the same way, and simply offset each element of that input vector by a fixed small value, using those offsets as start and end keys for a couch view query. I didn’t test this on larger vectors but have always been curious on how far it could go. Probably the matching algorithms in this project are more accurate, but this naive method worked pretty well for my use case and on a maintenance free DBaaS. Would love to hear more about concrete usage examples in this area if anybody likes to share - thanks Edit - found my slightly more detailed blog post about this http://splatcollision.com/page/fast-vector-similarity-queries-using-couchdb-views http://splatcollision.com/page/fast-vector-similarity-querie...