A company is using a legacy telephony platform and has several years remaining on its contract. The company wants to move to AWS and wants to implement the following machine learning features:
– Call transcription in multiple languages
– Categorization of calls based on the transcript
– Detection of the main customer issues in the calls
– Customer sentiment analysis for each line of the transcript, with positive or negative indication and scoring of that sentiment
Which AWS solution will meet these requirements with the LEAST amount of custom model training?

A. Use Amazon Transcribe to process audio calls to produce transcripts, categorize calls, and detect issues. Use Amazon Comprehend to analyze sentiment.
B. Use Amazon Transcribe to process audio calls to produce transcripts. Use Amazon Comprehend to categorize calls, detect issues, and analyze sentiment
C. Use Contact Lens for Amazon Connect to process audio calls to produce transcripts, categorize calls, detect issues, and analyze sentiment.
D. Use Contact Lens for Amazon Connect to process audio calls to produce transcripts. Use Amazon Comprehend to categorize calls, detect issues, and analyze sentiment.

A company is building an application that can predict spam email messages based on email text. The company can generate a few thousand human-labeled datasets that contain a list of email messages and a label of "spam" or "not spam" for each email message. A machine learning (ML) specialist wants to use transfer learning with a Bidirectional Encoder Representations from Transformers (BERT) model that is trained on English Wikipedia text data.
What should the ML specialist do to initialize the model to fine-tune the model with the custom data?

A. Initialize the model with pretrained weights in all layers except the last fully connected layer.
B. Initialize the model with pretrained weights in all layers. Stack a classifier on top of the first output position. Train the classifier with the labeled data.
C. Initialize the model with random weights in all layers. Replace the last fully connected layer with a classifier. Train the classifier with the labeled data.
D. Initialize the model with pretrained weights in all layers. Replace the last fully connected layer with a classifier. Train the classifier with the labeled data.

A company wants to deliver digital car management services to its customers. The company plans to analyze data to predict the likelihood of users changing cars. The company has 10 TB of data that is stored in an Amazon Redshift cluster. The company's data engineering team is using Amazon SageMaker Studio for data analysis and model development. Only a subset of the data is relevant for developing the machine learning models. The data engineering team needs a secure and cost-effective way to export the data to a data repository in Amazon S3 for model development.
Which solutions will meet these requirements? (Choose two.)

A. Launch multiple medium-sized instances in a distributed SageMaker Processing job. Use the prebuilt Docker images for Apache Spark to query and plot the relevant data and to export the relevant data from Amazon Redshift to Amazon S3.
B. Launch multiple medium-sized notebook instances with a PySpark kernel in distributed mode. Download the data from Amazon Redshift to the notebook cluster. Query and plot the relevant data. Export the relevant data from the notebook cluster to Amazon S3.
C. Use AWS Secrets Manager to store the Amazon Redshift credentials. From a SageMaker Studio notebook, use the stored credentials to connect to Amazon Redshift with a Python adapter. Use the Python client to query the relevant data and to export the relevant data from Amazon Redshift to Amazon S3.
D. Use AWS Secrets Manager to store the Amazon Redshift credentials. Launch a SageMaker extra-large notebook instance with block storage that is slightly larger than 10 TB. Use the stored credentials to connect to Amazon Redshift with a Python adapter. Download, query, and plot the relevant data. Export the relevant data from the local notebook drive to Amazon S3.
E. Use SageMaker Data Wrangler to query and plot the relevant data and to export the relevant data from Amazon Redshift to Amazon S3.

A machine learning (ML) specialist has prepared and used a custom container image with Amazon SageMaker to train an image classification model. The ML specialist is performing hyperparameter optimization (HPO) with this custom container image to produce a higher quality image classifier.
The ML specialist needs to determine whether HPO with the SageMaker built-in image classification algorithm will produce a better model than the model produced by HPO with the custom container image. All ML experiments and HPO jobs must be invoked from scripts inside SageMaker Studio notebooks.
How can the ML specialist meet these requirements in the LEAST amount of time?

A. Prepare a custom HPO script that runs multiple training jobs in SageMaker Studio in local mode to tune the model of the custom container image. Use the automatic model tuning capability of SageMaker with early stopping enabled to tune the model of the built-in image classification algorithm. Select the model with the best objective metric value.
B. Use SageMaker Autopilot to tune the model of the custom container image. Use the automatic model tuning capability of SageMaker with early stopping enabled to tune the model of the built-in image classification algorithm. Compare the objective metric values of the resulting models of the SageMaker AutopilotAutoML job and the automatic model tuning job. Select the model with the best objective metric value.
C. Use SageMaker Experiments to run and manage multiple training jobs and tune the model of the custom container image. Use the automatic model tuning capability of SageMaker to tune the model of the built-in image classification algorithm. Select the model with the best objective metric value.
D. Use the automatic model tuning capability of SageMaker to tune the models of the custom container image and the built-in image classification algorithm at the same time. Select the model with the best objective metric value.

A retail company uses a machine learning (ML) model for daily sales forecasting. The model has provided inaccurate results for the past 3 weeks. At the end of each day, an AWS Glue job consolidates the input data that is used for the forecasting with the actual daily sales data and the predictions of the model. The AWS Glue job stores the data in Amazon S3.
The company's ML team determines that the inaccuracies are occurring because of a change in the value distributions of the model features. The ML team must implement a solution that will detect when this type of change occurs in the future.
Which solution will meet these requirements with the LEAST amount of operational overhead?

A. Use Amazon SageMaker Model Monitor to create a data quality baseline. Confirm that the emit_metrics option is set to Enabled in the baseline constraints file. Set up an Amazon CloudWatch alarm for the metric.
B. Use Amazon SageMaker Model Monitor to create a model quality baseline. Confirm that the emit_metrics option is set to Enabled in the baseline constraints file. Set up an Amazon CloudWatch alarm for the metric.
C. Use Amazon SageMaker Debugger to create rules to capture feature values Set up an Amazon CloudWatch alarm for the rules.
D. Use Amazon CloudWatch to monitor Amazon SageMaker endpoints. Analyze logs in Amazon CloudWatch Logs to check for data drift.

An ecommerce company wants to train a large image classification model with 10,000 classes. The company runs multiple model training iterations and needs to minimize operational overhead and cost. The company also needs to avoid loss of work and model retraining.
Which solution will meet these requirements?

A. Create the training jobs as AWS Batch jobs that use Amazon EC2 Spot Instances in a managed compute environment.
B. Use Amazon EC2 Spot Instances to run the training jobs. Use a Spot Instance interruption notice to save a snapshot of the model to Amazon S3 before an instance is terminated.
C. Use AWS Lambda to run the training jobs. Save model weights to Amazon S3.
D. Use managed spot training in Amazon SageMaker. Launch the training jobs with checkpointing enabled.

A company's machine learning (ML) specialist is designing a scalable data storage solution for Amazon SageMaker. The company has an existing TensorFlow-based model that uses a train.py script. The model relies on static training data that is currently stored in TFRecord format.
What should the ML specialist do to provide the training data to SageMaker with the LEAST development overhead?

A. Put the TFRecord data into an Amazon S3 bucket. Use AWS Glue or AWS Lambda to reformat the data to protobuf format and store the data in a second S3 bucket. Point the SageMaker training invocation to the second S3 bucket.
B. Rewrite the train.py script to add a section that converts TFRecord data to protobuf format. Point the SageMaker training invocation to the local path of the data. Ingest the protobuf data instead of the TFRecord data.
C. Use SageMaker script mode, and use train.py unchanged. Point the SageMaker training invocation to the local path of the data without reformatting the training data.
D. Use SageMaker script mode, and use train.py unchanged. Put the TFRecord data into an Amazon S3 bucket. Point the SageMaker training invocation to the S3 bucket without reformatting the training data.

An automotive company is using computer vision in its autonomous cars. The company has trained its models successfully by using transfer learning from a convolutional neural network (CNN). The models are trained with PyTorch through the use of the Amazon SageMaker SDK. The company wants to reduce the time that is required for performing inferences, given the low latency that is required for self-driving.
Which solution should the company use to evaluate and improve the performance of the models?

A. Use Amazon CloudWatch algorithm metrics for visibility into the SageMaker training weights, gradients, biases, and activation outputs. Compute the filter ranks based on this information. Apply pruning to remove the low-ranking filters. Set the new weights. Run a new training job with the pruned model.
B. Use SageMaker Debugger for visibility into the training weights, gradients, biases, and activation outputs. Adjust the model hyperparameters, and look for lower inference times. Run a new training job.
C. Use SageMaker Debugger for visibility into the training weights, gradients, biases, and activation outputs. Compute the filter ranks based on this information. Apply pruning to remove the low-ranking filters. Set the new weights. Run a new training job with the pruned model.
D. Use SageMaker Model Monitor for visibility into the ModelLatency metric and OverheadLatency metric of the model after the model is deployed. Adjust the model hyperparameters, and look for lower inference times. Run a new training job.

A company is building a machine learning (ML) model to classify images of plants. An ML specialist has trained the model using the Amazon SageMaker built-in Image Classification algorithm. The model is hosted using a SageMaker endpoint on an ml.m5.xlarge instance for real-time inference. When used by researchers in the field, the inference has greater latency than is acceptable. The latency gets worse when multiple researchers perform inference at the same time on their devices. Using Amazon CloudWatch metrics, the ML specialist notices that the ModelLatency metric shows a high value and is responsible for most of the response latency.
The ML specialist needs to fix the performance issue so that researchers can experience less latency when performing inference from their devices.
Which action should the ML specialist take to meet this requirement?

A. Change the endpoint instance to an ml.t3 burstable instance with the same vCPU number as the ml.m5.xlarge instance has.
B. Attach an Amazon Elastic Inference ml.eia2.medium accelerator to the endpoint instance.
C. Enable Amazon SageMaker Autopilot to automatically tune performance of the model.
D. Change the endpoint instance to use a memory optimized ML instance.

A manufacturing company needs to identify returned smartphones that have been damaged by moisture. The company has an automated process that produces 2,000 diagnostic values for each phone. The database contains more than five million phone evaluations. The evaluation process is consistent, and there are no missing values in the data. A machine learning (ML) specialist has trained an Amazon SageMaker linear learner ML model to classify phones as moisture damaged or not moisture damaged by using all available features. The model's F1 score is 0.6.
Which changes in model training would MOST likely improve the model's F1 score? (Choose two.)

A. Continue to use the SageMaker linear learner algorithm. Reduce the number of features with the SageMaker principal component analysis (PCA) algorithm.
B. Continue to use the SageMaker linear learner algorithm. Reduce the number of features with the scikit-learn multi-dimensional scaling (MDS) algorithm.
C. Continue to use the SageMaker linear learner algorithm. Set the predictor type to regressor.
D. Use the SageMaker k-means algorithm with k of less than 1,000 to train the model.
E. Use the SageMaker k-nearest neighbors (k-NN) algorithm. Set a dimension reduction target of less than 1,000 to train the model.