Neural Networks, IEEE - INNS - ENNS International Joint Conference on (2000)
July 24, 2000 to July 27, 2000
F. Rosati , Universit? di Ancona
P. Campolucci , Universit? di Ancona
F. Piazza , Universit? di Ancona
A large class of non-linear dynamic adaptive systems such as dynamic recurrent neural networks can be very effectively represented by Signal-Flow-Graphs (SFGs). By this method, complex systems are described as a general connection of many simple components, each of them implementing a simple one-input one-output transformation, as in an electrical circuit. Even if graph representations are popular in the neural network community, they are often use d for qualitative description rather than for rigorous representation and computational purposes. Following an approach originally developed by A.Y. Lee for continuous-time systems based on the concept of adjoint graph, a new algorithm to estimate the derivative of the output with respect to an internal parameter was recently proposed by some of the authors for discrete-time systems. This paper extends further this approach to multirate digital systems, which are nowadays widely used. The new method can be employed for gradient-based learning of general multirate circuits, such as new “multirate” neural networks.
F. Piazza, F. Rosati and P. Campolucci, "A General Approach to Gradient Based Learning in Multirate Systems and Neural Networks," Neural Networks, IEEE - INNS - ENNS International Joint Conference on(IJCNN), Como, Italy, 2000, pp. 4569.