similarity can be generated offline, but prediction can be generated online on a per request basis. it eliminate latency, but has a little performance overhead.

Comments

danithaca’s picture

Status: Active » Closed (duplicate)