Concept icon
Concept

Choose evaluation data to match the real-world target

A model is often trained on one kind of data but ultimately needs to work well on a different kind of data. For example, a wildlife classifier might be trained mostly on clear camera-trap photos, but the real deployment goal is to recognize animals in blurry field-camera images taken at night. In that situation, the development and test sets should be built from the same kind of images the model will face after deployment, because those sets should reflect the future data distribution that matters most.

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

Related