Google Professional Machine Learning Engineer Question 142
Select 3Google Cloud PlatformYou are tasked with quickly prototyping a machine learning model for a customer churn prediction system using Jupyter notebooks on Google Cloud. Which considerations should you include to ensure the prototyping process is effective and scalable?
- A
Select a small but representative sample of the dataset for faster experimentation.
- B
Use managed Jupyter environments like Vertex AI Workbench to avoid setup overhead.
- C
Deploy the model to production immediately after completing prototyping in the notebook.
- D
Track and log hyperparameter settings and experiment results during prototyping.
- E
Focus solely on model architecture without considering data preprocessing during prototyping.
Show answer and explanation
Correct answers: A, B, D
Explanation
Effective model prototyping in Jupyter notebooks requires leveraging tools and techniques that enhance experimentation speed, reproducibility, and scalability. Using a representative dataset sample, managed environments like Vertex AI Workbench, and proper logging are essential for efficient prototyping. However, notebooks are not suitable for direct production deployment, and data preprocessing should not be ignored during prototyping as it impacts model performance.
- A. Correct.
Using a small but representative sample of the dataset allows for faster experimentation while still providing meaningful insights during prototyping.
- B. Correct.
Using managed environments like Vertex AI Workbench can save time and ensure the environment is scalable and integrated with other Google Cloud services.
- C. Incorrect.
Deploying a model to production directly from a Jupyter notebook is not recommended, as notebooks are intended for experimentation and lack the robustness required for production environments.
- D. Correct.
Tracking hyperparameter settings and experiment results ensures reproducibility and helps in evaluating the effectiveness of different configurations during prototyping.
- E. Incorrect.
While model architecture is important, ignoring data preprocessing can lead to unrealistic results, as preprocessing is a critical step in the ML pipeline.