Choose a dev set size that can detect a tiny but important gain.
Case context: A ranking team for an established movie-streaming platform cares about a 0.01% lift in accuracy because even that small change can noticeably affect subscription revenue. Its current development set has 10,000 examples.
Question: What should the team do about the development set, and what is the reason?
Sample answer: The team should strongly consider expanding the development set beyond 10,000 examples. When the goal is to notice an improvement as small as 0.01%, a set of only 10,000 examples may be too coarse to measure that change reliably. The right size is not chosen by a fixed rule; it should be large enough to make the desired improvement detectable, but not larger than necessary.
Key points:
- The team cares about a 0.01% improvement.
- Small but meaningful gains matter in mature systems.
- A 10,000-example set may not be sensitive enough here.
- The best size depends on the smallest change the team needs to observe.
Rubric: The response should recommend increasing the development set beyond 10,000 examples, explain that this is needed to detect a 0.01% improvement, and state that the set should be sized to match the meaningful change the team wants to measure.
0
1
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Tiny Development Sets Miss Very Small Accuracy Gains
Typical Development Set Sizes for Tiny Accuracy Gains
High-Impact Business Systems May Justify a Larger Development Set
Formal significance tests for validation-set changes
What dev set size is most suitable for spotting a 0.1 percentage-point gain in accuracy?
A development set should always be expanded to the maximum possible size, even after it is already large enough to reveal meaningful performance changes.
Validation set size for noticing a tiny accuracy change
Match each evaluation target with the dev set size it suggests.
Order the steps for deciding whether a dev set is large enough to detect a useful accuracy gain.
Match dev-set size to the smallest gain you care about.
Choose a dev set size that can detect a tiny but important gain.
Why is a 150-example dev set not enough to tell 83.0% from 83.4% accuracy?
When is a validation set much larger than 10,000 examples most justified?
If a validation set is already large enough to tell whether one model is meaningfully better than another, it does not need to be made much larger.