Google Professional Machine Learning Engineer Question 196
Single answerGoogle Cloud PlatformYou are a machine learning engineer working on a recommendation system for an e-commerce platform. You are tasked with running multiple experiments to determine the best model architecture and hyperparameters. You want to track each experiment’s metrics, configurations, and results systematically while maintaining reproducibility. Which approach should you use to achieve this using Google Cloud tools?
- A
Use Vertex AI Experiments to track and compare the metrics, configurations, and outputs of your experiments.
- B
Store all experiment configurations and results manually in a Cloud Storage bucket for later analysis.
- C
Use a custom-built logging system deployed on Compute Engine to track experiment results and metrics.
- D
Leverage Vertex AI Feature Store to track experiments and evaluate their performance.
Show answer and explanation
Correct answer: A
Explanation
Vertex AI Experiments is the best tool for tracking machine learning experiments on Google Cloud. It enables you to log experiment configurations, metrics, and results systematically, making it easier to compare different runs and maintain reproducibility. Other approaches, such as manually storing results in Cloud Storage or building custom solutions, are less efficient and do not provide the same level of integrated functionality.
- A. Correct.
Correct. Vertex AI Experiments is specifically designed for tracking, managing, and comparing experiments in machine learning workflows. It helps maintain reproducibility and provides insights into model performance.
- B. Incorrect.
Incorrect. While Cloud Storage can store experiment data, it lacks the systematic tracking, comparison tools, and reproducibility features offered by Vertex AI Experiments.
- C. Incorrect.
Incorrect. Building a custom logging system requires additional development and maintenance effort, and it may not provide the robust tracking and comparison capabilities available in Vertex AI Experiments.
- D. Incorrect.
Incorrect. Vertex AI Feature Store is designed for managing and serving features for machine learning models, not for tracking experiments or comparing results.