Amazon AWS Certified Machine Learning - Specialty Exam Practice Questions (P. 2)

GPT-4o - Answer

Collaborative filtering, as correctly indicated by the consensus among user comments, is indeed the most appropriate approach for this scenario where the goal is to identify customer shopping patterns, trends, and preferences based on user interactions and correlations. Understanding that LDA and Random Cut Forest target entirely different problem types like topic modeling and anomaly detection respectively, and neural networks, although versatile, would need more contextual tuning which makes it not the default choice here. Therefore, collaborative filtering stands out as it directly leverages user interaction data to generate personalized recommendations.

GPT-4o - Answer

For the dataset containing PII stored in Amazon S3, to ensure that it is accessible only from a VPC and doesn’t utilize the public internet, setting up a VPC endpoint is essential. Adding a specific bucket access policy, which effectively limits the accessibility to this VPC-only via the set VPC endpoint, addresses both security concerns. By doing this, no EC2 specifics are necessary nor is the use of ALLOW policies from various security peripherals like NACLs or different groups, ensuring a streamlined and secure access route that adheres strictly to the given constraints.

GPT-4o - Answer

Amazon Transcribe, Amazon Translate, and Amazon Comprehend together provide the most seamless path to achieving sentiment analysis from a Spanish video for an English speaker. First, Transcribe converts spoken language in the video into written Spanish. Then, Translate converts the Spanish text into English. Finally, Comprehend analyzes the English text to determine sentiment. This sequence provides a completely managed service experience without the need for additional coding or infrastructure setup, ensuring an efficient solution.

GPT-4o - Answer

Upon reviewing the latest AWS documentation as referenced by users in their comments, it becomes clear that the correct preparation for leveraging NVIDIA GPUs in a Docker container on Amazon EC2 P3 instances involves ensuring that the container is NVIDIA-Docker compatible. Importantly, specialists should not bundle NVIDIA drivers within the Docker image; instead, they should include only the CUDA toolkit. This adjustment aligns with best practices for configuring environments in AWS SageMaker for efficient GPU utilization during model training.