QuestionHardMetrics and A/B TestingML System Design question from post-interview feedback · T-BankT-Bank
Training pointwise-ranker without leaks and distortion in popular goods
You train a boosting-ranker for recommendations on clicks and bundles of images. How to collect a dataset, make a validation split and not retrain on popular products and old shows?
Answer as you would in an interview
Answer aloud or write down a few key points. The text field is optional; the review opens after your attempt.