Google Professional ML Engineer Practice Test

1. Which Google Cloud service is best for training large-scale machine learning models with minimal infrastructure management?

Vertex AI

Cloud Storage

BigQuery ML

Cloud Functions

2. What is the primary purpose of using cross-validation in model evaluation?

To reduce training time

To estimate model performance on unseen data

To increase dataset size

To normalize feature values

3. In BigQuery ML, which statement correctly creates a linear regression model?

CREATE MODEL my_model AS SELECT * FROM data;

CREATE OR REPLACE MODEL my_model OPTIONS(model_type='linear_reg') AS SELECT * FROM data;

TRAIN MODEL my_model FROM data;

BUILD MODEL my_model USING linear_reg;

4. What is the recommended approach for handling imbalanced datasets in classification tasks?

Ignore the minority class

Use oversampling, undersampling, or adjust class weights

Always use stratified sampling

Remove imbalanced data entirely

5. Which metric is most appropriate for evaluating a binary classification model with imbalanced classes?

Accuracy

Precision and Recall

Mean Squared Error

R-squared

6. What does hyperparameter tuning in Vertex AI Vizier optimize?

Model weights during training

Configuration parameters to maximize model performance

Data preprocessing steps

Feature engineering rules

7. Which preprocessing technique is essential before training a neural network with features of different scales?

One-hot encoding

Feature normalization or standardization

Categorical binning

Feature hashing

8. In the context of model deployment on Vertex AI, what is a canary deployment?

Deploying to all users simultaneously

Routing a small percentage of traffic to a new model version for testing

Rolling back to a previous model

Deploying during off-peak hours

9. Which Google Cloud service allows you to build and deploy custom ML models without writing code?

Cloud ML Engine

Vertex AI AutoML

TensorFlow

Cloud Dataflow

10. What is the purpose of regularization in machine learning models?

To increase training speed

To prevent overfitting and improve generalization

To reduce dataset size

To normalize input features

11. When should you use batch prediction instead of online prediction in Vertex AI?

When predictions are needed immediately

When processing large volumes of data with latency tolerance

For real-time applications

When model accuracy is critical

12. Which evaluation metric is best for regression problems?

Precision

F1-score

Mean Absolute Error (MAE) or Root Mean Squared Error (RMSE)

Confusion matrix

13. In machine learning pipelines, what is the primary role of feature engineering?

To clean raw data

To create relevant features that improve model performance

To split data into train and test sets

To deploy models to production

14. What is the benefit of using Vertex AI Pipelines for ML workflows?

Automatic hyperparameter tuning

Orchestration, reproducibility, and scalability of ML workflows

Real-time data processing

Automatic feature selection

15. Which approach is recommended for monitoring model performance in production?

Monitor only prediction latency

Track data drift, prediction drift, and model accuracy metrics continuously

Manually test the model weekly

No monitoring needed after deployment

Google Professional Machine Learning Engineer Practice Test