EpochLex entry
Backpropagation
bak-PROP-uh-GAY-shun
The algorithm used in neural-network training to calculate how changes in model parameters affect loss, allowing gradients to be propagated backward through the network so the parameters can be updated.

