Short Answer

Rationale for a 2,000-Example Eyeball Dev Set at a 5% Error Rate

Question: Briefly explain why an Eyeball dev set needs to be approximately 2,000 examples large when a classifier's error rate is exactly 5%.

Sample answer: At a 5% error rate, 2,000 examples are needed to ensure the Eyeball dev set contains about 100 misclassified examples (0.05 * 2,000 = 100), which provides a large enough set of errors to manually analyze.

Key points:

  • 0.05 * 2,000 = 100
  • The goal is to obtain roughly 100 misclassified examples for analysis

Rubric: The response should explicitly state the math (5% of 2000) or mention the target of capturing approximately 100 errors for analysis.

0

1

Updated 2026-06-07

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