Google Professional Machine Learning Engineer Question 68
Select 4Google Cloud PlatformYour team is working on a project to classify customer feedback into multiple categories such as 'Positive', 'Negative', and 'Neutral'. You want to use Google Cloud AutoML to build a text classification model with minimal manual intervention. However, your team has limited machine learning expertise. Which of the following steps are required to successfully train and deploy a model using AutoML Natural Language?
- A
Prepare a labeled dataset in a CSV file with columns for text and labels.
- B
Upload the labeled dataset to a Google Cloud Storage bucket.
- C
Manually design and optimize the architecture of the neural network.
- D
Use the AutoML Natural Language interface to import your dataset and train the model.
- E
Download and deploy the trained model locally on your own servers for inference.
- F
Evaluate the model using AutoML's built-in tools to assess its performance.
Show answer and explanation
Correct answers: A, B, D, F
Explanation
Google Cloud AutoML Natural Language is designed to simplify the process of building and deploying machine learning models, especially for users with limited ML expertise. By preparing and uploading a labeled dataset, using the AutoML interface, and leveraging built-in tools for training and evaluation, you can quickly build a text classification model. Manual architecture design and local deployment are unnecessary since AutoML automates these processes and hosts the model on Google Cloud.
- A. Correct.
Correct: AutoML Natural Language requires a labeled dataset, typically in CSV format, where you provide the text data and the corresponding labels.
- B. Correct.
Correct: To use the dataset in AutoML, it must first be uploaded to a Google Cloud Storage bucket.
- C. Incorrect.
Incorrect: AutoML handles the model architecture and optimization automatically, so no manual intervention is needed for this step.
- D. Correct.
Correct: The AutoML Natural Language interface allows you to import your dataset, train the model, and manage the process easily without requiring deep ML expertise.
- E. Incorrect.
Incorrect: AutoML models are deployed on Google Cloud infrastructure. You can use REST APIs to access the model, but deploying it locally is not part of the AutoML workflow.
- F. Correct.
Correct: AutoML provides built-in evaluation tools, such as confusion matrices and precision-recall metrics, to help assess the model's performance.