Case Study

Redirecting an ML Team's Target Alignment Mid-Project

Case context: A machine learning team starts a project with an initial dev set and evaluation metric to ensure they can iterate rapidly. Six weeks into the project, the team realizes their metric and dev set no longer reflect the actual product goals or user needs, meaning their models are optimizing for the wrong target.

Question: Based on the principles in Machine Learning Yearning, how should the team handle this misalignment regarding their dev/test sets and metrics, and what final step is critical once they make this decision?

Sample answer: The team should not hesitate to change their dev/test sets and evaluation metrics since changing them mid-project is common and not a big deal when they no longer point the team in the right direction. Once the change is decided, the critical final step is to make sure the entire team knows about the new direction.

Key points:

  • Change the misaligned dev/test sets or evaluation metrics.
  • Understand that changing metrics/sets is a common and acceptable practice.
  • Ensure the team is informed about the new direction.

Rubric: Grading should verify that the student: 1. Advises changing the dev/test sets or metrics since they are misaligned. 2. Identifies this action as a common practice that is 'not a big deal'. 3. Specifies that the team must be notified of the new direction.

0

1

Updated 2026-05-26

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Related