Neural network simulator for circuit modeling and analysis based on fast automatic differentiation

In this paper, we propose a neural network simulator (NNS) based on fast automatic differentiation (FAD) for the circuit modeling and analysis. By the proposed NNS the circuits having the complex nonlinear and piecewise linear (PWL) functions are modeled and analyzed. The NNS with a state variable v adds the FAD object of the differentiation for a function f(v) as its differential elements in the Jacobian matrix. This means that the user defines only the description of many nonlinear functions in the input file. Similarly, in the NNS piecewise linear function p(v) is combined as a voltage controlling current source. The simulations and the results are shown for their analysis.