Circuit theoretic approach to the backpropagation learning algorithm
Giovanni Martinelli, R. Perfetti · 1991
The widespread use of multilayer perceptrons in applications is due to the availability of a simple and efficient learning algorithm such as the backpropagation learning (BPL) algorithm. The authors point out the circuit theoretic basis of BPL by showing that it can be directly derived by the adjoint network method applied to the perceptron represented by the corresponding electrical circuit. This algorithm, within the framework of circuit theory is useful. Its usefulness is discussed.>