Short Answer

Why is a 150-example dev set not enough to tell 83.0% from 83.4% accuracy?

Question: Answer in one to three sentences: Why is a 150-example dev set not enough to distinguish classifiers with 83.0% and 83.4% accuracy?

Sample answer: The gap is only 0.4 percentage point, but each mistake on a 150-example set changes accuracy by about 0.67 percentage point. That makes the estimate too noisy to reliably tell those models apart.

Key points:

  • The difference is only 0.4 percentage point.
  • One mistake changes accuracy by about 0.67 point on 150 examples.
  • A larger dev set is needed to compare them reliably.

Rubric: The answer must identify the small 0.4 percentage-point difference and state that 150 examples is too small, or too noisy, to detect it.

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