Google Professional Machine Learning Engineer Question 264
Single answerGoogle Cloud PlatformYou are training a machine learning model on Google Cloud AI Platform to predict customer churn. The model has several hyperparameters, such as learning rate, batch size, and the number of layers. You want to optimize these hyperparameters to improve model performance. Which of the following strategies should you use to efficiently tune the hyperparameters while minimizing computational cost?
- A
Use Grid Search to exhaustively explore all possible combinations of hyperparameter values.
- B
Use Random Search to randomly sample different combinations of hyperparameter values.
- C
Use Bayesian Optimization with Vertex AI Vizier to intelligently explore the hyperparameter space.
- D
Manually adjust hyperparameter values based on trial-and-error.
Show answer and explanation
Correct answer: C
Explanation
Bayesian Optimization with Vertex AI Vizier is specifically designed to optimize hyperparameters efficiently by intelligently exploring the search space. It reduces computational cost by focusing on the most promising areas, unlike Grid Search or Random Search, which lack such prioritization. This makes it the most suitable approach for hyperparameter tuning in production machine learning workflows on Google Cloud.
- A. Incorrect.
Grid Search exhaustively evaluates all combinations of hyperparameters, which can be computationally expensive and inefficient, especially for large search spaces.
- B. Incorrect.
Random Search is more efficient than Grid Search but still lacks the capability to prioritize promising areas of the hyperparameter space.
- C. Correct.
Bayesian Optimization with Vertex AI Vizier intelligently explores the hyperparameter space by leveraging prior evaluations to focus on the most promising regions, making it an efficient and cost-effective solution.
- D. Incorrect.
Manually adjusting hyperparameter values is time-consuming, error-prone, and not scalable for complex models.