Training ANN is tricky as it contains several parameters to optimize. The procedure of updating the weights is called backpropagation. The procedure to minimize the error is called optimization.