Task 2.2: Train and Refine Models MLA-C01 Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   Which factor directly influences the number of parameters in a fully connected neural network layer?

   • AThe batch size used during training
   • BThe activation function used in the layer
   • CThe learning rate of the optimizer
   • DThe number of input features and the number of neurons in the layer
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   What is the primary purpose of dropout as a regularization technique?

   • ATo add noise to the training labels to make the model more robust
   • BTo permanently remove unimportant weights from the model after training
   • CTo increase the learning rate during training to speed up convergence
   • DTo randomly deactivate a fraction of neurons during training to reduce overfitting
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A data scientist has trained a custom scikit-learn model locally and wants to deploy it to SageMaker for real-time inference. The model is saved as a joblib file. What is the correct sequence of steps to deploy this model?

   • AUpload the joblib file directly to SageMaker Studio and use the built-in deployment wizard without an inference script.
   • BConvert the model to a SageMaker built-in algorithm format and then deploy it.
   • CPackage the model with an inference script, upload the artifacts to Amazon S3, create a SageMaker Model, and deploy it to an endpoint.
   • DRegister the model in SageMaker Model Registry first, then deploy it directly from the registry without creating a SageMaker Model.
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   Which SageMaker feature allows you to track multiple versions of a model and manage their approval status?

   • AModel Registry
   • BAutomatic Model Tuning
   • CData Wrangler
   • DExperiments
   Reveal Solution Discussion Select an answer first
6. 30foundation · easy

   When bringing a custom model into SageMaker for deployment, what must you provide in addition to the model artifacts?

   • AAn inference script or a custom container that defines how to load the model and serve predictions.
   • BA SageMaker Ground Truth labeling job.
   • CA new training dataset to validate the model.
   • DAn AWS Lambda function for every prediction request.
   Reveal Solution Discussion Select an answer first