Short Answer

Resampling Rule and Curves for Reducing Variability

Question: If you repeatedly build many tiny training samples to make a noisy performance curve easier to read, what kind of resampling should be used for those samples, and what two mean error curves should be shown?

Sample answer: Those subsets should be formed by bootstrap-style resampling, so individual examples can appear more than once in the same subset. The two curves to plot are the mean training error and the mean development-set error.

Key points:

  • Create the small subsets by resampling with replacement.
  • Show the average training error and the average development-set error.

Rubric: The response must state that the subsets are drawn by resampling with replacement and must name both the mean training error and the mean development-set error (or average dev error) as the plotted quantities.

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