Multiple ChoiceBeginner 45s
Recommendation candidate generation stage
In a large-scale recommender, why is candidate generation separated from ranking?
Options
- 1To cheaply narrow millions of items to a few hundred before expensive scoring
- 2Because the ranking model cannot practically score more than a handful of items at once
- 3To avoid storing any feature vectors for the items in the candidate pool at all
- 4To eliminate the need to run A/B tests on the final ranking stage of the pipeline
Think you can answer under pressure?
Get matched and answer live questions in 3-minute 1v1 battles. Elo ranking, badge tiers and bragging rights on the line.