Case Study

Prioritizing Pipeline Components in a Mammogram Classifier

Case context: You are developing a two-stage medical imaging pipeline to detect tumors. The first component performs tumor detection, and the second component classifies the detected tumors. An analysis reveals that the tumor detection component is performing far below human-level performance, while the classification component is performing very close to human-level performance.

Question: Based on the principle of comparing pipeline components to human-level performance, which component should your team focus on improving and why?

Sample answer: The team should focus on improving the tumor detection component. Because its performance is far from human-level, it represents a clear opportunity where significant improvements are possible. In contrast, the classification component is already performing close to human-level, meaning further optimization will likely be highly challenging and yield minimal returns.

Key points:

  • Identify the tumor detection component as the priority for improvement.
  • Explain that the tumor detection component is far from human-level performance, which indicates room for significant improvement.
  • State that the classification component is close to human-level performance and therefore a lower priority due to the difficulty of making further gains.

Rubric: The response must correctly identify the tumor detection component as the priority. It must justify this choice by explaining that a component far from human-level performance has a strong case for improvement, whereas a component close to human-level performance is harder to improve.

0

1

Updated 2026-05-27

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI