Skip to content

Model Training & Adaptation

Train-test split

Keeping some examples separate from AI learning so they can provide an independent check afterward.

Example

A team teaches AI with one collection and checks it with another.

Why people use it

It gives AI a fairer check on examples kept apart from its learning.

What you'll hear

“Keep these records aside until the final check.”

What this means for you

Keep closely related or duplicate examples from making the final check artificially easy.

Can you control it?

Developer-only

The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.

Common questions

Does any random split create a fair test?
No. Related records, duplicates or time dependence can cross the boundary.
Why can records from the same person cause trouble?
Similar information on both sides can make the test easier than predicting for a new person.
Should future records usually be used to predict the past?
No. A time-based task should reflect what information would actually have been available then.

Related terms

Still have questions?

Up to 500 characters.

Ask LATHIC about AI. Relevant glossary entries may be included.

Your question, the glossary entries it matches, and a rotating pseudonymous identifier go to Microsoft Azure’s OpenAI service through Vercel AI Gateway to generate an answer. Zero retention and no training are required of the provider, and LATHIC does not save your question or answer. Privacy Notice