Databricks Machine Learning Professional Question 202
Select 2A retail company is deploying a recommendation system to suggest products to users as they browse the website. The company wants the system to deliver predictions immediately after a user interacts with the website to enhance the user experience. Why would real-time inference be advantageous in this scenario?
- A
It minimizes latency, providing immediate predictions to improve user experience.
- B
It reduces the computational cost compared to batch inference.
- C
It is well-suited for scenarios requiring predictions for a small number of records at a time.
- D
It eliminates the need for any model deployment infrastructure.
- E
It allows continuous retraining of the model during inference.
Show answer and explanation
Correct answers: A, C
Explanation
Real-time inference is particularly beneficial in scenarios where low latency is critical and predictions are needed for a small number of records at a time. In this use case, providing immediate recommendations after user interactions enhances the user experience and aligns with the goals of the retail company's system. However, real-time inference typically requires robust deployment infrastructure and incurs higher computational costs compared to batch processing.
- A. Correct.
Correct: Real-time inference is designed to minimize latency, making it ideal for applications like recommendation systems that require instant predictions.
- B. Incorrect.
Incorrect: Real-time inference is generally more computationally expensive than batch inference due to the need to process requests individually.
- C. Correct.
Correct: Real-time inference shines when predictions are needed for a small number of records at a time, as in user-specific recommendations.
- D. Incorrect.
Incorrect: Real-time inference requires model deployment infrastructure such as REST APIs or streaming services to handle live requests.
- E. Incorrect.
Incorrect: Real-time inference does not involve retraining models during inference; retraining is a separate process typically performed offline.