Answer-first summary for fast verification
Answer: Incorporate images of the new products into your test dataset as soon as they are added to the retraining process, ensuring the continuous evaluation service has current data to assess model performance accurately.
Option A is correct because it ensures that the continuous evaluation service can accurately assess the model's performance on both existing and new products. By updating the test dataset with new product images as they are introduced to the retraining process, the service can evaluate the model's ability to adapt to new data while maintaining high accuracy across all products. This approach supports the goal of continuous improvement and adaptation of the ML model.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
No comments yet.
In your role as a Machine Learning Engineer at an online retail company, you are tasked with enhancing a visual search engine. The company is about to launch a new line of products, and you've configured an end-to-end ML pipeline on Google Cloud to detect if an image contains the company's product. To accommodate the new products, you've enabled the retraining feature in the pipeline. Utilizing AI Platform's continuous evaluation service, you aim to ensure the models maintain high accuracy on your test dataset. Considering the need for the model to adapt to new products without compromising the evaluation of existing product detection, what is the BEST approach to manage your test dataset? (Choose one correct option)
A
Incorporate images of the new products into your test dataset as soon as they are added to the retraining process, ensuring the continuous evaluation service has current data to assess model performance accurately.
B
Wait to update your test dataset with images of the new products until the evaluation metrics indicate a drop below a predefined threshold, to avoid unnecessary data changes.
C
Maintain your test dataset unchanged, relying on the retraining process alone to adapt the model to new products without evaluating their detection separately.
D
Replace your entire test dataset with images of the new products once they are introduced to retraining, to focus evaluation solely on the new products' detection.