we run this stream processing in Recommender Systems. It has at least two advances reading from streaming:
1. fast feedback to recall model. the model should have the ability to know the results for given recommended object, so streaming feedback is suitable here. 2. easy scaling. when the deployed ML models for prediction have reached the max throughput, a new instance will be scaled up which should read the same dataset from streaming store.
Thanks pod...@gmx.com wrote:
What’s the best system structure for ‘stream processing’ in such situations? 😊
-- Simple Mail https://simplemail.co.in/
