Google Professional Machine Learning Engineer
Get started today
Ultimate access to all questions.
Ultimate access to all questions.
Your company has recently migrated its machine learning infrastructure to Google Cloud, transitioning from disk-based block storage to Cloud Storage for file storage and BigQuery for tabular data, with Dataflow handling pre-processing. The team primarily uses TensorFlow and Keras for model development. Given this setup, which of the following methods is NOT suitable for accessing tabular data from BigQuery as part of your machine learning pipeline? Choose the best option.
Explanation:
The correct answer is D because tf.data.Iterator is designed for iterating over elements within a TensorFlow Dataset and cannot directly access or iterate over BigQuery tables without prior data loading or transformation. Option A is suitable as the BigQuery I/O Connector is designed for use with Dataflow to process data directly from BigQuery. Option B is appropriate because the BigQuery Storage API allows efficient reading of data into TensorFlow models. Option C is valid for scenarios where data exploration or lightweight processing is needed before model training. Option E represents a best practice by combining initial data exploration with scalable processing, making it also correct. Therefore, when E is an option, both D and E are correct answers, with D being the primary answer as it directly addresses the unsuitability for the described scenario.