An Oscillatory Deep Neural Network for Coupled Electrical Circuits

Abstract Electronic systems share an indispensable role in almost every modern industry and are therefore continuously evolving into more advanced and complex versions. Consequently, such systems need to be tackled with some cutting-edge techniques. Among a number of analytical and numerical techniques of this era, Artificial Neural Networks (ANNs) have grabbed attention due to their universality and robustness on assigned tasks. In this work, an oscillatory Deep Neural Network (DNN) model has been proposed with an oscillatory activation function and specific layers’ structure to learn the dynamics of coupled LC-series circuits. The DNN model being suggested is flexible, easy to implement, and capable of diligently recovering the vibrating patterns of underlying dynamical systems. Outputs from the network are being compared with the results of LSODA numerical solvers. An error analysis for different time spans has also being performed, validating the successful recovery of solutions to the modeled problem, which is evident to the competency of proposed technique.