Split Data Between Endpoints for Realtime Interference MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A manufacturing company deploys a quality-control model to a real-time endpoint. The training data includes sensor readings from machines A and B. The endpoint serves data from machine C, which has different sensor characteristics. What is the most appropriate action to maintain model accuracy?

   • AUse the model as-is because it was trained on similar machines.
   • BRoute all traffic to a secondary endpoint that uses a different model.
   • CIncrease the endpoint's compute resources to handle the new data.
   • DRetrain the model on data from all three machines.
   Reveal Solution Discussion Select an answer first
2. 7expert · hard

   A healthcare company deploys a patient-risk model to a real-time endpoint. The training data was collected from hospital A, but the endpoint serves data from hospital B. The data distributions differ significantly. What is the most appropriate action to maintain model accuracy?

   • ARetrain the model on a combined dataset from both hospitals.
   • BRoute all traffic to a secondary endpoint that uses a different model.
   • CIncrease the endpoint's compute resources to handle the new data.
   • DUse the model as-is because it was trained on similar data.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   When monitoring a real-time inference endpoint, what should you adjust if you observe a decline in accuracy?

   • AThe number of requests sent to the endpoint
   • BThe data split between training and serving data
   • CThe hardware specifications of the serving infrastructure
   • DThe latency threshold of the endpoint
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A ride-sharing company deploys a fare-prediction model to a real-time endpoint. They monitor the endpoint's accuracy and latency. After a major city event, they observe a spike in latency but accuracy remains stable. What is the most appropriate action?

   • ARetrain the model to adapt to the event data.
   • BScale up the endpoint's compute resources to handle the increased load.
   • CAdjust the data split to send more traffic to a secondary endpoint.
   • DIgnore the latency spike as it is temporary.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   What is the primary purpose of separating training data from the data used to serve real-time inference requests?

   • ATo ensure the model is evaluated on data it has never seen during training
   • BTo allow the training data to be updated more frequently than the serving data
   • CTo reduce the storage cost of the training dataset
   • DTo simplify the deployment pipeline by using the same data for both training and validation
   Reveal Solution Discussion Select an answer first