Case Study

Diagnosing a team's learning curve trend

Case context: A machine learning team has plotted a learning curve for their dev-set error. They started with 500 examples and observed a dev-set error of 25%. They subsequently increased the training set to 2,000 examples, then 5,000 examples, and are analyzing the resulting dev-set errors.

Question: Based on the fundamental relationship described in Machine Learning Yearning, what trend must the team observe in the dev-set error as they evaluate the model trained on 2,000 and 5,000 examples compared to the initial 500?

Sample answer: The team should observe a decreasing trend in the dev-set error. As they move from 500 to 2,000 and then to 5,000 training examples, the dev-set error should steadily drop from the initial 25%.

Key points:

  • The dev-set error should decrease.
  • This expected decrease occurs as the training set size increases.

Rubric: The student must correctly identify that the dev-set error is expected to decrease as the training set size grows from 500 to 5,000 examples.

0

1

Updated 2026-05-27

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