Databricks Machine Learning Associate Question 78
Select 3You are working on a machine learning project in Databricks and need to store preprocessed features for reuse across multiple models. Which steps should you follow to create a feature store table in Databricks?
- A
Use the Databricks Feature Store client to create a feature table and specify its schema.
- B
Write the feature data directly to a Delta table without using the Feature Store API.
- C
Register the feature table in the Feature Store by calling the
create_tablemethod. - D
Ensure the feature table is associated with a Databricks MLflow experiment for versioning and tracking.
- E
Use a Feature Store API method to compute and write features from a DataFrame to the feature table.
Show answer and explanation
Correct answers: A, C, E
Explanation
To create a feature store table in Databricks, you must use the Databricks Feature Store client to define and register the feature table. This involves specifying the schema and using the create_table method to register the table in the Feature Store. Additionally, you need to use a Feature Store API method to compute and write features from a DataFrame to the registered table. Writing directly to a Delta table bypasses the Feature Store's functionality, and associating the table with an MLflow experiment is optional and not part of the creation process.
- A. Correct.
Correct. Using the Databricks Feature Store client and defining the schema is a key step in creating a feature table.
- B. Incorrect.
Incorrect. Writing data directly to a Delta table bypasses the Feature Store API, meaning the table won't be registered or usable as a feature table.
- C. Correct.
Correct. Registering the feature table with the
create_tablemethod is essential to make it available for reuse in the Feature Store. - D. Incorrect.
Incorrect. Associating a feature table with an MLflow experiment is not required to create the table, although MLflow can be used for model tracking.
- E. Correct.
Correct. Using the Feature Store API to compute and write features ensures proper integration and management of feature data.