I am learning about neural networks and specially back propagation.
How can one show in case a neural network appears more than once that the gradient of the loss function according to this weight matrix is a sum of two gradients? For an example, please refer to example 3.