Google Professional Machine Learning Engineer Question 266
Select 3Google Cloud PlatformYou are training a machine learning model for predicting product sales using Google Cloud's AI Platform. The training process involves several hyperparameters such as learning rate, batch size, and the number of layers in a neural network. You want to optimize these hyperparameters to improve the model's performance. Which of the following steps should you take to efficiently perform hyperparameter tuning on Google Cloud?
- A
Use Vertex AI's Hyperparameter Tuning service to define a hyperparameter tuning job.
- B
Manually adjust hyperparameters by trial and error and re-run training jobs until the best configuration is found.
- C
Define a range of hyperparameter values and use random search to sample combinations during training.
- D
Use Vertex AI to define a search space for hyperparameters and specify a Bayesian optimization strategy.
- E
Run multiple training jobs in parallel using preemptible VM instances to reduce costs.
Show answer and explanation
Correct answers: A, D, E
Explanation
To optimize hyperparameters efficiently on Google Cloud, you should leverage the tools provided by Vertex AI, such as the Hyperparameter Tuning service. Defining a search space and using Bayesian optimization is more effective than random search or manual tuning. Additionally, running parallel jobs using preemptible instances reduces costs while speeding up the tuning process.
- A. Correct.
Correct. Vertex AI's Hyperparameter Tuning service is specifically designed for automated and efficient hyperparameter tuning, making it a recommended approach.
- B. Incorrect.
Incorrect. Manually adjusting hyperparameters is time-consuming, error-prone, and inefficient compared to automated methods available in Google Cloud.
- C. Incorrect.
Incorrect. Random search may work but is less efficient compared to advanced optimization strategies like Bayesian optimization provided by Vertex AI.
- D. Correct.
Correct. Using Vertex AI's Bayesian optimization strategy helps in efficiently exploring the hyperparameter space to find the best configuration.
- E. Correct.
Correct. Running parallel training jobs on preemptible VM instances is a cost-efficient way to speed up hyperparameter tuning.