Essay

Why does choosing one evaluation metric improve model selection?

Question: A product team is comparing several image classifiers for a quality-control system. Suppose the team has no agreed evaluation metric and instead debates each candidate model by looking at them one by one. Analyze what problems this causes. Then explain how defining a new trusted metric helps the team move forward.

Sample answer: Without a trusted metric, the team ends up comparing models by hand, which is slow and often inconsistent. Different people may favor different models for different reasons, so the group lacks a single objective target. Once the team defines one trusted metric, everyone can optimize toward the same goal. That makes model comparison automatic, reduces debate, and lets the team iterate much faster.

Key points:

  • Without a trusted metric, model choice becomes a manual process.
  • Manual comparison is slow and can produce mixed or subjective decisions.
  • A trusted metric gives the team one clear objective for progress.

Rubric: Response should explain both the drawbacks of manual model selection and the advantage of using a trusted metric to unify evaluation and speed development.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Machine Learning Strategy

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Yearning @ DeepLearning.AI