Google Professional Machine Learning Engineer Question 400
Single answerGoogle Cloud PlatformYour organization is implementing a hybrid cloud strategy to train machine learning models. Sensitive data must remain on-premises, but the organization wants to leverage Google Cloud's TPU for faster training. Which solution allows you to securely train models in this hybrid setup while ensuring data privacy?
- A
Use Vertex AI and transfer the sensitive data to Google Cloud for training.
- B
Use TensorFlow Federated to perform federated learning across on-premises and Google Cloud.
- C
Use Google Cloud Interconnect to establish a secure connection and train the model entirely on-premises.
- D
Use Anthos to deploy a hybrid ML pipeline that processes data on-premises and trains models in Google Cloud.
Show answer and explanation
Correct answer: D
Explanation
Anthos is the best solution for this hybrid strategy because it allows organizations to maintain sensitive data on-premises while leveraging Google Cloud's capabilities, such as TPUs, for model training. This approach meets the requirements for data privacy and efficient model training in a hybrid cloud environment.
- A. Incorrect.
This option does not satisfy the data privacy requirement, as sensitive data would be transferred to Google Cloud.
- B. Incorrect.
TensorFlow Federated is designed for federated learning, a decentralized approach to training, but it is not the best fit for this specific hybrid strategy where sensitive data remains on-premises, and training occurs in the cloud.
- C. Incorrect.
This option incorrectly suggests using Google Cloud Interconnect to train models entirely on-premises, which does not leverage Google Cloud's TPU for training.
- D. Correct.
Anthos enables hybrid and multicloud deployments, allowing you to process sensitive data on-premises while offloading training tasks to Google Cloud TPU. This satisfies both the data privacy and performance requirements.