An ML engineer is using Amazon QuickSight anomaly detection to detect very high or very low machine operating temperatures compared to normal. The ML engineer sets the Severity parameter to Low and above. The ML engineer sets the Direction parameter to All.
What effect will the ML engineer observe in the anomaly detection results if the ML engineer changes the Direction parameter to Lower than expected?
A. Increased anomaly identification frequency and increased recall
B. Decreased anomaly identification frequency and decreased recall
C. Increased anomaly identification frequency and decreased recall
D. Decreased anomaly identification frequency and increased recall
Show Answer
Correct Answer: B
Explanation: Changing Direction from All to Lower than expected limits detection to only negative (lower-than-expected) anomalies. High anomalies are no longer flagged, so the total number of detected anomalies decreases. Relative to detecting both high and low anomalies, this also reduces recall because true high anomalies are intentionally excluded.
Question 42
A company is using an Amazon SageMaker AI ML model to predict traffic accidents that potholes cause. An ML engineer has configured SageMaker Model Monitor to run as part of a SageMaker AI pipeline. In the MonitoringExecution output, the ML engineer observes several baseline_drift_check violations that are failing the pipeline.
What should the ML engineer do to resolve this issue?
A. Retrain the model by using a new SageMaker AI training job. Check for errors by using SageMaker Debugger.
B. Retrain the model with new training data. Reuse the original baseline in Model Monitor.
C. Retrain the model with new training data. Use the new baseline in Model Monitor.
D. Rerun the SageMaker AI pipeline after enabling the emit_metrics option in the baseline constraints file.
Show Answer
Correct Answer: C
Explanation: A baseline_drift_check violation indicates the production data distribution has drifted from the baseline statistics and constraints. The appropriate remediation is to retrain the model using representative new training data and generate a new baseline (statistics/constraints) from that updated data for Model Monitor. Reusing the old baseline would continue to flag the new distribution, Debugger is unrelated to data drift, and enabling emit_metrics does not resolve drift violations.
Question 43
An ML engineer is building a model to predict house and apartment prices. The model uses three features: Square Meters, Price, and Age of Building. The dataset has 10,000 data rows. The data includes data points for one large mansion and one extremely small apartment.
The ML engineer must perform preprocessing on the dataset to ensure that the model produces accurate predictions for the typical house or apartment.
Which solution will meet these requirements?
A. Remove the outliers and perform a log transformation on the Square Meters variable.
B. Keep the outliers and perform normalization on the Square Meters variable.
C. Remove the outliers and perform one-hot encoding on the Square Meters variable.
D. Keep the outliers and perform one-hot encoding on the Square Meters variable.
Show Answer
Correct Answer: A
Explanation: Removing the two extreme outliers (the unusually large mansion and extremely small apartment) helps the model focus on the typical housing distribution when the goal is accurate predictions for typical properties. Square Meters is a continuous numeric feature, and a log transformation is an appropriate preprocessing step to reduce skew and compress large values. One-hot encoding is intended for categorical variables, not continuous measurements like square meters. Normalization alone does not address the influence of extreme outliers.
Question 44
A bank needs to use Amazon SageMaker AI to create an ML model to determine which customers qualify for a new product. The bank must use algorithms that SageMaker AI directly supports. The model must be explainable to the bank’s regulators.
Which modeling approach will meet these requirements?
A. Train the model by using the Object2Vec algorithm.
B. Train the model by using the linear learner algorithm.
C. Train a neural network.
D. Train the model by using the k-means algorithm.
Show Answer
Correct Answer: B
Explanation: The problem is a supervised binary classification task (qualify or not qualify). Amazon SageMaker Linear Learner is a built-in algorithm that supports binary classification and produces linear models whose feature weights are interpretable, making them suitable for explainability requirements from regulators. Object2Vec is for representation learning, neural networks are generally less explainable, and k-means is an unsupervised clustering algorithm, not appropriate for qualification classification.
Question 45
A company is using Amazon SageMaker AI to develop a credit risk assessment model. During model validation, the company finds that the model achieves 82% accuracy on the validation data. However, the model achieved 99% accuracy on the training data. The company needs to address the model accuracy issue before deployment.
Which solution will meet this requirement?
A. Add more dense layers to increase model complexity. Implement batch normalization. Use early stopping during training.
B. Implement dropout layers. Use L1 or L2 regularization. Perform k-fold cross-validation.
C. Use principal component analysis (PCA) to reduce the feature dimensionality. Decrease model layers. Implement cross-entropy loss functions.
D. Augment the training dataset. Remove duplicate records from the training dataset. Implement stratified sampling.
Show Answer
Correct Answer: B
Explanation: The large gap between training accuracy (99%) and validation accuracy (82%) indicates overfitting. Dropout and L1/L2 regularization are standard techniques to reduce overfitting by limiting model reliance on specific weights. K-fold cross-validation provides a more robust estimate of generalization performance during validation. The other options either increase model complexity (A), include techniques not primarily aimed at this issue or mix less relevant changes (C), or focus on data handling without directly addressing the observed overfitting (D).
Question 46
HOTSPOT
-
A company needs to combine data from multiple sources. The company must use Amazon Redshift Serverless to query an AWS Glue Data Catalog database and underlying data that is stored in an Amazon S3 bucket.
Select and order the correct steps from the following list to meet these requirements. Select each step one time or not at all. (Select and order three.)
• Attach the IAM role to the Redshift cluster.
• Attach the IAM role to the Redshift namespace.
• Create an external database in Amazon Redshift to point to the Data Catalog schema.
• Create an external schema in Amazon Redshift to point to the Data Catalog database.
• Create an IAM role for Amazon Redshift to use to access only the S3 bucket that contains underlying data.
• Create an IAM role for Amazon Redshift to use to access the Data Catalog and the S3 bucket that contains underlying data.
Show Answer
Correct Answer: Step 1: Create an IAM role for Amazon Redshift to use to access the Data Catalog and the S3 bucket that contains underlying data.
Step 2: Attach the IAM role to the Redshift namespace.
Step 3: Create an external schema in Amazon Redshift to point to the Data Catalog database.
Explanation: Amazon Redshift Serverless uses namespaces (not clusters) for IAM role attachment. The IAM role needs permissions for both AWS Glue Data Catalog and the underlying Amazon S3 data. Redshift Spectrum integrates with the Glue Data Catalog by creating an external schema that references the catalog database.
Question 47
HOTSPOT
-
An ML engineer needs to automate the rebuild and redeployment of an ML model. Updates will occur when changes are made to the model’s code base. The ML engineer must use AWS services to configure a continuous integration and continuous delivery (CI/CD) pipeline for the rebuild and redeployment.
Select and order the steps from the following list to configure the CI/CD pipeline. Each step should be selected one time. (Select and order three.)
• Invoke Amazon SageMaker Pipelines to run all steps required for model training and deployment.
• Create a pipeline in AWS CodePipeline. Build and test containers in AWS CodeBuild.
• Create a Git source code repository.
Show Answer
Correct Answer: Step 1: Create a Git source code repository.
Step 2: Create a pipeline in AWS CodePipeline. Build and test containers in AWS CodeBuild.
Step 3: Invoke Amazon SageMaker Pipelines to run all steps required for model training and deployment.
Explanation: The CI/CD flow starts with a source repository, then CodePipeline/CodeBuild detect changes, build and test artifacts, and finally trigger SageMaker Pipelines to retrain and deploy the model.
Question 48
A company runs its ML workflows on an on-premises Kubernetes cluster. The ML workflows include ML services that perform training and inferences for ML models. Each ML service runs from its own standalone Docker image.
The company needs to perform a lift and shift from the on-premises Kubernetes cluster to an Amazon Elastic Kubernetes Service (Amazon EKS) cluster.
Which solution will meet this requirement with the LEAST operational overhead?
A. Redesign the ML services to be configured in Kubeflow. Deploy the new Kubeflow managed ML services to the EKS cluster.
B. Upload the Docker images to an Amazon Elastic Container Registry (Amazon ECR) repository. Configure a deployment pipeline to deploy the images to the EKS cluster.
C. Migrate the training data to an Amazon Redshift cluster. Retrain the models from the migrated training data by using Amazon Redshift ML. Deploy the retrained models to the EKS cluster.
D. Configure an Amazon SageMaker AI notebook. Retrain the models with the same code. Deploy the retrained models to the EKS cluster.
Show Answer
Correct Answer: B
Explanation: A lift-and-shift migration from an on-premises Kubernetes cluster to Amazon EKS should reuse the existing containerized ML services with minimal changes. Uploading the existing Docker images to Amazon ECR and deploying them to EKS preserves the current architecture and minimizes operational overhead. The other options require redesigning the platform (Kubeflow), changing the ML workflow (Redshift ML), or retraining and moving to SageMaker notebooks, all of which add unnecessary effort and operational complexity.
Question 49
A company wants to migrate ML models from an on-premises environment to Amazon SageMaker AI. The models are based on the PyTorch algorithm. The company needs to reuse its existing custom scripts as much as possible on AWS.
Which feature of SageMaker AI should the company use to meet these requirements?
A. SageMaker AI built-in algorithms
B. SageMaker Canvas
C. SageMaker JumpStart
D. SageMaker AI script mode
Show Answer
Correct Answer: D
Explanation: SageMaker AI script mode is designed to let you bring your own training and inference scripts while using SageMaker-managed framework containers such as PyTorch. This enables migration of existing custom PyTorch code with minimal changes. Built-in algorithms require using SageMaker-provided algorithms, Canvas is a no-code tool, and JumpStart provides pretrained models and solution templates rather than reusing existing custom scripts.
Question 50
An ML engineer is training a text generation model on Amazon SageMaker AI. After several epochs, the loss function does not converge, and the model’s accuracy on the validation dataset starts to show oscillating results.
The ML engineer needs to ensure that the model achieves generalization.
Which solution will meet this requirement?
A. Increase the learning rate and decrease the mini-batch size.
B. Increase the learning rate as the number of epochs increases.
C. Decrease the learning rate and increase the mini-batch size.
D. Decrease the learning rate and decrease the mini-batch size.
Show Answer
Correct Answer: C
Explanation: A non-converging loss with oscillating validation accuracy commonly indicates the learning rate is too high and/or gradient estimates are too noisy. Reducing the learning rate leads to more stable optimization, while increasing the mini-batch size reduces gradient variance, improving convergence and supporting better generalization.
$19
Get all 235 questions with detailed answers and explanations
Instant download HTML + PDF delivered the moment payment clears.
Secure Stripe checkout we never see or store your card details.
7-day refund if files are defective see our refund policy.
