Short Answer

How to Spot High Variance from a Gap in Errors

Question: Suppose a model has about 0% attainable error, its training error is 1%, and its training-dev error is 5%. Which error measurement shows that the model is suffering from high variance?

Sample answer: The 5% error on the training-dev set is the signal of high variance.

Key points:

  • Training error is only 1%.
  • Training-dev error rises to 5%.
  • The larger gap points to a variance problem.

Rubric: The answer must explicitly state the 5% training-dev error rate.

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