Think Together · Start with a Short Story
The girl gives different importance to two clues: “has fur” and “pointy ears.” The robot reminds her that weights decide how each input affects a calculation.
Lay Out the Process
- InputA clue value
- WeightMultiply by a factor
- CombineCalculate with other inputs
- TrainingAdjust the factors
A weight is an adjustable parameter in a model that affects how an input influences later calculations.
A positive, negative, or near-zero weight can have different effects. In a network with many layers, one number alone cannot tell you that it represents a particular human concept.
Training updates these parameters based on targets and errors. The model does not simply “like” one clue more than another.
Now You Try
If an input is 4 and its weight is 0, how much does it add to this weighted sum?
Ready? Open this to see the explanation
It adds 0. But the whole network may still use related information through another path, so one connection does not tell the whole story.
One Idea to Take Away
A weight is an adjustable parameter in a model that affects how an input influences later calculations.
Try covering the diagram above and explaining each step to someone in your family. If one step is hard to explain, return to that card and look again.