Discuss ML Project Tradeoffs

In a senior machine learning interview, you are asked to walk through a previous recommendation or prediction project in detail. Pick one specific project as the basis and respond to these technical decision-making questions:

1. What led you to select the modeling approach you used?
2. Why did you avoid a more complex model?
3. How would you balance precision, recall, latency, and infrastructure cost?
4. What would you do if offline metrics and online business metrics conflict?
5. How would you tune the decision threshold of a binary classifier?
6. Compare a Factorization Machine to an embedding-based deep neural network for recommendation or ranking. Overview: This question assesses a candidate's ability in model selection; balancing precision, recall, latency, and infrastructure cost; tuning thresholds; reconciling offline and online metrics; and comparing recommendation models in the machine learning domain.