Model Training & Adaptation
Backpropagation
The algorithm used to calculate gradients through a neural network.
Example
Error signals move backward through the network to update earlier layers.
Why it matters
It makes training large networks practical by sharing the work of calculating learning signals.
You'll hear it
“The error calculation needs to reach the earlier parts of the network.”
Can you control it?
Set by the team
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
- Is it the same thing as the final prediction?
- No. It is part of learning how to adjust the system, rather than the ordinary act of using it.
- Does it decide which examples are suitable?
- No. The teaching material and learning target must still be chosen appropriately.
- Can it be used with different update methods?
- Yes. Its calculated learning signals can guide several kinds of adjustment methods.