AI Glossary

Backpropagation

Also called backprop

What does “Backpropagation” mean?

Backpropagation: The algorithm that works out how much each weight in a neural network contributed to the error. Starting from the output, it works backward through the layers using the chain rule from calculus, and it produces the gradients that gradient descent needs.

Backpropagation visual guide: The algorithm that works out how much each weight in a neural network contributed to the error. Starting from the output, it works backward through the layers using the chain rule from calculus, and it produces the gradients that gradient descent needs.
Visual guide Open full size (new tab)

What is an example of “Backpropagation”?

Without backpropagation, training a network with billions of weights would be impractical, because computing each weight's effect separately would take far too long.

Keep learning

See also

Gradient descentTermLoss functionTermNeural networkTermLearning rateTerm

Same topic

Activation functionTermBatch sizeTermCheckpointTermComputeTermCurriculum learningTerm