Essay

Explain How Error Category Frequency Guides Project Prioritization

Question: Explain how determining the frequency of different error categories (such as Dog, Great Cat, or Blurry images) in a misclassified dev set helps a machine learning team decide which projects to prioritize. Provide a specific example based on Andrew Ng's framework.

Sample answer: Determining error category frequency provides an upper bound on how much a specific fix can improve performance. If an error analysis on 100 dev set examples shows that 'Dog' mistakes account for only 8% of the errors, resolving all dog-related issues will eliminate at most 8% of total errors. In contrast, if 'Great Cat' or 'Blurry' images account for a larger percentage of errors, the team should prioritize those categories, as working on them could help eliminate more errors overall.

Key points:

  • Error category frequency establishes the maximum possible improvement (upper bound) for fixing that specific error type.
  • Fixing 'Dog' mistakes can eliminate at most 8% of the errors.
  • Categories with higher frequencies ('Great Cat' or 'Blurry' images) should be prioritized because they offer the potential to eliminate more total errors.

Rubric: The essay should explain that frequency sets a ceiling on potential improvement and correctly apply the example of prioritizing Great Cat/Blurry image errors over Dog errors because they offer a higher ceiling.

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