Model Training & Adaptation
Bagging
Combining predictions from several AI systems taught using different samples drawn from the same collection.
Example
Several prediction systems give estimates that are averaged together.
Why people use it
It can make predictions less dependent on one particular set of teaching examples.
What you'll hear
“Let's combine several versions instead of trusting just one.”
What this means for you
Check the combined result on examples none of the systems learned from.
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 bagging create new real-world evidence?
- No. Resampling reuses the available observations.
- Does every version see exactly the same examples?
- No. The method uses resampled collections, so some examples appear more often and others are omitted.
- Can all the versions share the same weakness?
- Yes. Combining them does not fix a problem that affects them all.