Advanced Feature Store Concepts MACHINE-LEARNING-PROFESSIONAL Practice Questions (Page 4)

Questions 16–19

1. 16expert · hard

   A company has a feature table that is updated by a streaming pipeline every few seconds. They want to serve these features via an online table for real-time inference. They configure the online table using the Databricks SDK. However, they notice that the online table is not reflecting the latest feature values immediately. What is the most likely reason?

   • AThe online table has a synchronization lag, as it is updated asynchronously from the Delta table.
   • BThe online table is not compatible with streaming updates.
   • CThe online table is configured with the wrong primary key.
   • DThe streaming pipeline is not writing to the Delta table correctly.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   What is the primary purpose of the Feature Lookup API in Databricks Feature Store?

   • ATo deploy models to online serving endpoints.
   • BTo retrieve feature values for a given entity and timestamp, enabling consistent feature computation for training and inference.
   • CTo create and manage feature tables in the Unity Catalog.
   • DTo compute aggregate features from raw data in real time.
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A team is building a churn prediction model. They have a feature that is computed from raw customer data at inference time, such as 'number of support tickets opened in the last 7 days'. This feature is not stored in a feature table. How should they serve this feature consistently for both training and inference?

   • AUse a batch job to compute the feature and store it in a Delta table.
   • BStore the raw data in a feature table and compute the feature in the model code.
   • CUse on-demand features with feature serving to compute the feature at request time.
   • DSkip the feature because it is not stored.
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   In a training pipeline, how does the Feature Lookup API ensure that feature values are consistent with the inference environment?

   • ABy ensuring that all feature values are computed from the latest data available in the feature store.
   • BBy using the current timestamp for training data and the event timestamp for inference data.
   • CBy computing features from raw data at the time of training and storing them in the model artifact.
   • DBy using the same feature table and lookup logic for both training and inference, with timestamps representing the event time.
   Reveal Solution Discussion Select an answer first