Ultimate access to all questions.
As a Google Professional Machine Learning Engineer, you are tasked with enhancing a visual search engine for an online retail company. The ML pipeline is designed to classify images based on the presence of the company's products. With the anticipation of new product launches, a retraining feature has been integrated into the pipeline. To ensure the model maintains high accuracy, AI Platform's continuous evaluation service will be utilized. Considering the need for a robust evaluation strategy that accommodates new products without compromising the assessment of existing product classifications, what is the most effective approach to manage your test dataset? Choose the best option.
A
Extend your test dataset with images of the newer products as they are introduced, ensuring the evaluation covers both existing and new products.
B
Replace your test dataset entirely with images of the newer products upon their introduction, focusing evaluation solely on new product classifications.
C
Maintain the original test dataset without any updates, relying on the model's initial training to generalize to new products.
D
Monitor the model's performance metrics and only update the test dataset with images of newer products if the metrics fall below a predefined threshold.