Under what condition should you retrain a model with updated datasets?

• A. Never; retraining is never needed
• B. If the dataset size remains the same
• C. If the model is small and feature weights are updated
• D. If the model is large and heavy

Answer: (C)

Retraining with updated data is most appropriate when you can actually incorporate the new information into the model without excessive cost or downtime. In practice, a smaller model is ideal because you can quickly adjust and re-estimate its parameters, especially the feature weights, so the model quickly reflects new patterns in the data.

If the model is large and heavy, retraining becomes resource-intensive and time-consuming, making it less practical unless the update is crucial. The other statements aren’t reliable guides on their own: dataset size staying the same doesn’t imply retraining isn’t needed if data distributions have shifted, and never retraining ignores the need to adapt to new information.