1. You have been tasked with creating an Amazon EMR cluster for processing large-scale genomic datasets. The data is stored in Amazon S3, and the workflow will be run using Apache Spark. Which of the following configurations is the most appropriate for this use case?

A) A cluster with 1 master node and 10 core nodes, each with 32 vCPUs and 128 GB of memory, running on r4.4xlarge instances with EBS-optimized instances.
B) A cluster with 1 master node and 10 core nodes, each with 16 vCPUs and 64 GB of memory, running on r5d.2xlarge instances with EBS-optimized instances.
C) A cluster with 1 master node and 10 core nodes, each with 4 vCPUs and 16 GB of memory, running on t3.small instances with no EBS-optimized instances.
D) A cluster with 1 master node and 10 core nodes, each with 4 vCPUs and 16 GB of memory, running on m4.large instances with no EBS-optimized instances.
E) A cluster with 1 master node and 10 core nodes, each with 8 vCPUs and 32 GB of memory, running on c5.xlarge instances with no EBS-optimized instances.


2. In AWS Identity and Access Management (IAM), what is the recommended approach for granting machine learning models access to S3 buckets that store training data?

A) Use AWS Lambda functions to periodically download training data from the S3 bucket to the local file system of the machine learning model instance.
B) Create an IAM user for each model and provide it with access keys. Then, configure the model to use these access keys to access the S3 bucket.
C) Grant public access to the S3 bucket and configure the machine learning models to directly access the S3 bucket using an HTTP URL.
D) Create an IAM role for each model and attach a policy that allows read access to the specific S3 bucket. Then, specify the IAM role ARN in the AWS Glue job or SageMaker notebook instance that trains the model.



3. Which of the following is a disadvantage of using a small learning rate during model training in machine learning?

A) It can lead to slow convergence to the optimal solution
B) None of the above
C) It can lead to underfitting of the training data
D) It can cause the model to overfit the training data
E) It can result in unstable training due to large weight updates


4. What is the purpose of the AWS DeepLens device in machine learning workflows?

A) It is a cloud-based service that provides pre-trained models for image and video analysis.
B) It is a service that automatically creates and trains machine learning models using AWS data sets.
C) It is a device that provides real-time inference of machine learning models at the edge.
D) It is a device that provides high-performance computing for deep learning training.
E) It is a device that provides real-time data streaming for machine learning models.


5. A company has a large dataset of customer reviews stored in an Amazon S3 bucket. They want to use Amazon QuickSight to visualize the data and analyze the sentiment of the reviews. Which of the following options describes the best way to achieve this?

A) Use AWS Glue to crawl and catalog the data in S3, and then create a Glue ETL job to transform and load the data into Amazon Redshift. Connect QuickSight to Redshift and create the dashboard. Use Amazon Comprehend to analyze the sentiment of the reviews.
B) Use Amazon Kinesis Data Firehose to transform and load the data into Amazon Elasticsearch. Connect QuickSight to Elasticsearch and create the dashboard. Use Amazon Comprehend to analyze the sentiment of the reviews and set up alerts based on certain conditions.
C) Use AWS Lambda to process the data in S3, transform it, and load it into Amazon Athena. Connect QuickSight to Athena and create the dashboard. Use Amazon Comprehend to analyze the sentiment of the reviews and create custom metrics. Use these metrics to visualize the data in QuickSight.
D) Use the built-in connectors in QuickSight to connect to S3 and create a dashboard. Use Amazon Comprehend to analyze the sentiment of the reviews and create custom metrics. Use these metrics to visualize the data in QuickSight.