educative.io

Educative

Can you explain how to retrain the model every few hours per day

Hi Team, in the follow up question section, it said “Retrain the model as frequently as possible. One example is to retrain the model every few hours with new data (collected from user clicked data).”

Does this mean we just rerun the pre-train model (using historical data) multiple times per day using updated variables? not real retrain the model?


Course: Machine Learning System Design - Learn Interactively
Lesson: Ads Recommendation System Design - Machine Learning System Design

Hi @Kk_l, Thank you for contacting us.

For any prediction system to be accurate, the data on which the model is trained must be recent and updated, if we don’t frequently update our model then it will not adapt to the changing behavior of users, this is what we don’t want at any case because the user’s interest keeps on changing and so our model should predict accordingly.
These lines in the lesson mean that we should collect recent data every few hours and then retrain the model with the new data collected so that it makes predictions based on user’s recent interests and likes.

Hope Educative has helped you in your learning :slight_smile:

Thank @Noor_Awan, so it is actual retrain the model, not predict on the new data. Then how can we proceed this? Auto-ML?