Short Answer

Contrast the dev/test set distribution strategy for application progress versus research progress.

Question: According to Machine Learning Yearning, how do the recommended strategies for dev and test set distributions differ when the goal is progress on a specific ML application versus when the goal is general research progress?

Sample answer: For progress on a specific machine learning application, dev and test sets should be drawn from the same distribution to make the team more efficient. In contrast, research progress focuses on developing algorithms that are trained on one distribution and generalize well to a different distribution.

Key points:

  • Application progress requires dev and test sets from the same distribution.
  • Research progress focuses on generalizing algorithms from one distribution to another.
  • Same-distribution dev and test sets make the development team more efficient.

Rubric: The answer should state that application progress requires same-distribution dev/test sets for team efficiency, while research progress involves developing algorithms that generalize from one distribution to another.

0

1

Updated 2026-05-26

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Machine Learning Strategy

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Related