Learn Before
Concept icon
Concept

How to Select Dev and Test Data for the Future Task

Dev and test sets should be built from examples that match the data you expect to face after deployment, because those are the examples you want the model to do well on. A simple random split of the data you currently have can be misleading if future data will differ from the training examples. It is better to choose evaluation data based on the target use case than to assume the training and test distributions will naturally be the same.

0

1

Concept icon
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

Learn After