AI Foundations
Generalization
A model's ability to perform well on new examples it did not see during training.
Example
A model trained on several cities also predicts demand accurately in a city it never saw.
Why it matters
It makes earlier learning useful beyond the particular examples a system practiced on.
You'll hear it
“It worked in the pilot, but will it work for our other locations?”
Can you control it?
Not directly
No direct control. This describes a wider issue, concept or result rather than something you can simply switch on or off in a tool.
Common questions
- Does success on one new example establish reliability?
- No. A single success cannot show how the system handles the full variety of future cases.
- How do teams test generalization?
- They check examples kept apart from learning, including different groups, periods or conditions where possible.
- Why can a model fail on new data?
- The earlier examples may be too narrow, or changing conditions may make familiar patterns less useful.