Ask HN: How do you perform online learning for Production ML systems?
My workflow for batch/offline training takes a day. I have a use case where I want to aggregate feedback from last hour and perform online learning.
How would you do it in Production ML system serving live traffic? To answer this well, there are a few additional details needed:
1. How often does the model need to be retrained?
2. What's the output of your model?
3. You say training takes a day - if you had the parameters, how long does it take to process?