Google Professional Machine Learning Engineer Question 103
Select 4Google Cloud PlatformYou are designing a machine learning pipeline for a client that needs to process multiple types of data: tabular customer information, text-based customer reviews, and product images. The goal is to create a single model that can leverage all of this data for a recommendation system. How should you organize and prepare the data to ensure it is compatible with your machine learning model on Google Cloud?
- A
Normalize the tabular data and encode categorical features before feeding it into the model.
- B
Tokenize the text data and convert it into embeddings.
- C
Store all data types together in a single BigQuery table for simplicity.
- D
Preprocess image data by resizing and normalizing pixel values before inputting it into the model.
- E
Combine preprocessed tabular, text, and image data into a unified format, such as a TensorFlow Dataset.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
When dealing with multiple data types (tabular, text, images) in a single pipeline, each data type must be preprocessed according to its specific requirements to make it compatible with the machine learning model. Tabular data needs normalization and encoding; text data requires tokenization and embedding generation; image data must be resized and normalized. Once preprocessed, all data types should be combined into a unified format, such as a TensorFlow Dataset, to facilitate training a multi-modal model. Simply storing all data types together in a BigQuery table without preprocessing or conversion is insufficient for this purpose.
- A. Correct.
Tabular data often requires normalization to ensure numerical features are on a similar scale and encoding for categorical features to convert them into numerical representations that the model can understand.
- B. Correct.
Text data needs to be tokenized (split into smaller meaningful units) and converted into embeddings (numerical vectors) to make it usable for machine learning models.
- C. Incorrect.
Storing all data types in a single BigQuery table is not a practical solution as it does not address preprocessing or format compatibility for machine learning models.
- D. Correct.
Image data typically needs to be resized and normalized (e.g., pixel values scaled to [0,1]) to ensure consistency and compatibility with the model's input requirements.
- E. Correct.
Combining the preprocessed data from all modalities (tabular, text, image) into a unified format, such as a TensorFlow Dataset, is essential for training a single multi-modal model.