Google Professional Machine Learning Engineer Question 328
Select 3Google Cloud PlatformYou are designing a machine learning application that predicts customer churn for a subscription service. The application requires real-time feature updates and low-latency feature retrieval. You decide to use Vertex AI Feature Store to manage your features. Which of the following configurations and actions are appropriate for achieving these requirements?
- A
Enable online serving for the feature store to support low-latency feature retrieval.
- B
Store all features only in the offline store to simplify data management.
- C
Use a unique entity ID to identify each customer in the feature store.
- D
Leverage batch ingestion for real-time feature updates.
- E
Configure a feature monitoring job to ensure data quality in the feature store.
Show answer and explanation
Correct answers: A, C, E
Explanation
Vertex AI Feature Store provides an online store for low-latency, real-time feature retrieval and an offline store for batch processing. To meet the application's requirements, enabling online serving is critical for real-time predictions, and using a unique entity ID ensures efficient feature retrieval. Additionally, feature monitoring helps maintain data quality and consistency, which is vital for the success of the application.
- A. Correct.
Correct: Enabling online serving in Vertex AI Feature Store is necessary for low-latency feature retrieval, which is critical for real-time predictions.
- B. Incorrect.
Incorrect: Storing features only in the offline store does not support real-time or low-latency requirements. The offline store is designed for batch processing and training use cases.
- C. Correct.
Correct: A unique entity ID (e.g., customer ID) is required to uniquely identify and retrieve features from the feature store for each entity.
- D. Incorrect.
Incorrect: Batch ingestion is not suitable for real-time feature updates. Streaming ingestion or direct API calls are more appropriate for real-time use cases.
- E. Correct.
Correct: Configuring a feature monitoring job ensures the quality and consistency of data in the feature store, which is essential for reliable predictions.