Case Study

Case Study: Designing a cat detection dev/test set before mobile application launch.

Case context: You are building a mobile application to detect cats in user photos, but the app has not yet launched. Because you have no users, you do not have any actual user data to construct your dev and test sets.

Question: Based on the concept of approximating future dev/test data, how should you proceed to build your dev/test sets, and what is a specific action you can take to collect data?

Sample answer: You should proceed by trying to approximate the future user data distribution. To do this, you can ask your friends to take mobile phone pictures of cats and send them to you, which serves as a proxy to build the initial dev/test sets.

Key points:

  • Dev/test sets should be constructed pre-launch by approximating future data.
  • Proxy data collection is required since actual user data is unavailable.
  • Asking friends to take mobile-phone pictures of cats is a valid approximation method.

Rubric: The response must identify the need to approximate the future user data distribution and propose the specific action of asking friends to take mobile phone pictures of cats.

0

1

Updated 2026-05-26

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Machine Learning Strategy

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Related