AIF-C01 Question 43
Single answerA company wants to implement a machine learning solution for image classification but has limited resources to train a custom model from scratch. Which of the following approaches would be most suitable for their use case?
- A
Use an open-source pre-trained model and fine-tune it for their specific dataset.
- B
Build a custom machine learning model using a large-scale dataset collected internally.
- C
Leverage AWS services like Amazon Rekognition to perform image classification without requiring model training.
- D
Train a custom deep learning model using a cloud-based GPU cluster.
Show answer and explanation
Correct answer: A
Explanation
The best approach for a company with limited resources is to use an open-source pre-trained model and fine-tune it for their dataset. This method leverages existing work, saving time and computational cost, while still allowing for customization. Other options involve significant resource requirements or lack the flexibility to tailor the solution to the company’s specific needs.
- A. Correct.
This is correct. Using an open-source pre-trained model and fine-tuning it is a resource-efficient solution that allows customization for specific datasets while leveraging pre-existing training.
- B. Incorrect.
This is incorrect. Building a custom model from scratch requires significant time, expertise, and computational resources, which the company lacks based on the scenario.
- C. Incorrect.
This is incorrect. While Amazon Rekognition is a powerful AWS service, it is a fully managed service that does not allow for model customization, which may not meet the company’s need for fine-tuning.
- D. Incorrect.
This is incorrect. Training a custom deep learning model using a GPU cluster is resource-intensive and is not aligned with the company’s limited resources.