AI Foundations
Bias term
A learned offset that helps a model fit patterns that do not pass through zero.
Example
A prediction starts slightly above zero even when every measured input is zero.
Why it matters
It gives calculations flexibility to fit patterns that a fixed starting point would miss.
You'll hear it
“The prediction needs a starting adjustment, not just changes to the other clues.”
Can you control it?
Not directly
No direct control. A bias term is a learned offset inside a model. It comes from training rather than from anything a reader can adjust.
Common questions
- Is a bias term the same as AI bias?
- No. This is a number used in a calculation; unfair treatment is a different meaning of bias.
- Does it always have a positive value?
- No. Its value may be positive, negative or zero, depending on the learning setup.
- Does every AI design use it in the same way?
- No. Designs differ in where such adjustments appear and whether particular parts include them.