Short Answer

What problem is ruled out?

Question: A model has 8% error on the training set, 9% error on a training-development set, and 16% error on the dev set. This pattern points to high bias and a mismatch between the training source and the dev data. What issue is not suggested by these numbers?

Sample answer: It is not showing high variance on data from the same training distribution.

Key points:

  • The training and training-development errors are close, so variance is not the main concern.
  • The larger dev error points to a data mismatch rather than a variance problem.

Rubric: The answer must state that the model does not have high variance, especially on the training-data distribution.

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