Example

Accuracy Resolution on a 100-Example Development Set

For ordinary unweighted accuracy on 100 examples, each additional correct prediction changes the measured accuracy by 1 percentage point. The set therefore cannot directly register a 0.1-percentage-point difference such as 88.0% versus 88.1%, and sampling variation makes such a tiny underlying gain unreliable to infer from only 100 cases.

0

1

Updated 2026-08-30

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