Build indexes based on filter statistics #6
Description
When there are only a small set of points containing filters, we might want to build an LSH structure for just these points for fast lookup of points. This is not viable for 2 filters at a time, as we have 200k filters. However, if a query contains one small and one large filter (determined by how much they are used in data points), then we will query the smallest LSH structure of points and filter based on the second query metadata constraint.
We need to investigate whether this way of building indexes and querying based on filter statistics is the most efficient.