Short Definition
Backpropagation is the algorithm used to compute gradients through a neural network efficiently.
Intuition
It traces responsibility for the final error backward through the network so each weight knows how it should change.
Technical Definition
Backpropagation applies the chain rule of calculus through a computational graph to compute derivatives of the loss with respect to model parameters.
Example
After a classifier predicts the wrong class, backpropagation computes how each layer contributed to the loss.
Common Misunderstandings
Backpropagation computes gradients; an optimizer decides how to use them.
It is not specific to one neural network architecture.