Essay

When to Add a Mislabeled Category to Error Analysis

Question: In an error-analysis table, what should you do if you think annotation mistakes may be common, and what six column headings should the table contain?

Sample answer: If the amount of bad labeling appears large enough to matter, add a Mislabeled column so you can record how many examples have incorrect labels. The table should include these six headings: Image, Dog, Great cat, Blurry, Mislabeled, and Comments.

Key points:

  • Add a Mislabeled category when you suspect label errors are frequent enough to affect the analysis.
  • Use it to track the share of examples with incorrect annotations.
  • The six headings are Image, Dog, Great cat, Blurry, Mislabeled, and Comments.

Rubric: A complete answer must state the trigger for adding the category, explain that its role is to measure the fraction of mislabeled examples, and list all six table columns exactly as given: Image, Dog, Great cat, Blurry, Mislabeled, and Comments.

0

1

Updated 2026-08-12

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