Op-amp structures of fixed weight applications of BP neural network

Op-amp based scheme design of a fixed weight application of the backpropagation neural network is presented in this work. In this design, the most important limitations in op-amp utilization are investigated for design purposes. These limitations played an important role to obstruct the traditional learning phase. To encompass these limitations, sl-CONE model with its powerful learning algorithm is used, and to control the design aspect thoroughly, genetic algorithm is contributed efficiently. Within the genetic algorithm framework, a proper fitness function is designed and implemented. Weight results that are determined, match the design requirements. Unlike the present trends of similar works, the current work has calibrated the learning procedure in convenience with device characteristics instead of exploding complex schemes for the sake of responding to the learning requirements where, the computation appears as approximations of neural network operation results from the non-ideal characteristic signal bounding of physical devices.