Case Study

What conclusion is justified after a team reviews 20 Eyeball dev mistakes?

Case context: A team manually reviews about 20 mistakes made by its classifier on Eyeball dev examples. The review suggests that several error sources account for many of the observed mistakes.

Question: What should the team conclude from this review, using only the claim supported by the source?

Sample answer: The team can conclude that it is starting to gain a rough sense of the classifier's major error sources. It should not describe the review as a complete or exact account, because the source supports only a rough initial sense after about 20 mistakes.

Key points:

  • Manual review of about 20 mistakes
  • Eyeball dev examples
  • Major error sources
  • Rough rather than complete understanding

Rubric: Full credit requires a decision limited to a rough sense of major error sources and recognition that the evidence does not support claiming complete or exact knowledge.

0

1

Updated 2026-07-20

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Machine Learning Strategy

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Yearning @ DeepLearning.AI