Short Answer

Hypothesis Validation from Training to Training Dev Subset Performance

Question: When both the training and training dev sets contain car audio, what specific finding about the performance on these car audio subsets would validate the hypothesis that collecting more car audio data will help?

Sample answer: The hypothesis is validated if the system performs well on the car audio in the training set but performs poorly on the car audio in the training dev set.

Key points:

  • System does well on the car data subset in the training set.
  • System does poorly on the car data subset in the training dev set.

Rubric: The response should state that the model performs well/does well on the car data in the training set, but does not perform well/fails on the car data in the training dev set.

0

1

Updated 2026-05-26

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Related