Feature Store Management MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A team is building a feature store for a mobile app that personalizes content. They need to update user features in real-time as users interact with the app. Which type of feature table should they use for these updates?

   • AA temporary view that reads from a streaming source.
   • BOnline feature table with low-latency writes.
   • COffline feature table with hourly updates.
   • DOffline feature table with daily batch updates.
   Reveal Solution Discussion Select an answer first
2. 12expert · medium

   A company is building a real-time recommendation system. They need to serve features with low latency and also support batch training. They plan to use both online and offline feature tables. Which statement correctly describes the relationship between these tables?

   • AOnline tables are populated from offline tables through a synchronization process.
   • BOnline and offline tables are independent and can diverge in content.
   • CBoth tables must be updated in the same transaction.
   • DOffline tables are populated from online tables in real-time.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   Which of the following best describes an online feature table?

   • AA table optimized for low-latency reads, typically used for real-time inference.
   • BA table that stores historical feature values for batch processing.
   • CA table that is automatically replicated across multiple cloud regions.
   • DA table that is only accessible through the Databricks workspace UI.
   Reveal Solution Discussion Select an answer first
5. 14expert · medium

   A company has a model that is deployed for real-time inference. The model uses features from an online feature table. During a traffic spike, the team notices that some requests return stale features. What is the most likely cause?

   • AThe online table has eventual consistency.
   • BThe online table is not synchronized with the offline table.
   • CThe model is using the wrong feature table.
   • DThe feature table has too many partitions.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   What is the purpose of using the Feature Store API when scoring a model?

   • ATo automatically retrain the model with the new data.
   • BTo retrieve the latest feature values for the entities being scored, ensuring consistency with training.
   • CTo optimize the model's hyperparameters for better accuracy.
   • DTo store the scoring results back into the feature store.
   Reveal Solution Discussion Select an answer first