Backpropagation (machine learning)

Backpropagation in supervised machine learning is the process used to calculate the gradient of the error function associated with each parameter weighting within a convoluted neural network (CNN). Essentially, the gradient estimates how the system parameters should change in order to optimize the network overall .