backpropagation

Backpropagation is a widely-used algorithm for training neural networks that computes gradients of the loss function with respect to each weight by applying the chain rule, facilitating efficient updates during training.

19 papers