4.2 Key Issues in Neural Model Development

The preliminary step toward developing a neural model is the identification of the inputs and outputs of the problem to be modeled, so that the corresponding training data can be generated. In general, the model outputs are determined based on the purpose of the model for example, real and imaginary parts of S-parameters for linear circuit components, currents and charges for large-signal component models (harmonic balance), and cross-sectional RLCG parameters for high-speed VLSI interconnect analysis. Other factors influencing the choice of outputs are ease of data generation, ease of incorporation of the neural model into circuit simulators, and so forth. The neural model inputs are the device/circuit parameters that affect the outputs for example, circuit design variables, physical/process parameters of the components, and independent parameters such as frequency in passive components and bias for active components. As an illustration, to develop a small-signal FET model, we choose x = [ V _GS V _DS freq] ^T and y = [ RS ₁₁ IS ₁₁ RS ₁₂ IS ₂₂] ^T. Here, V _GS and D _DS, are gate-to-source and drain-to-source voltages, and RS _ij and IS _ij are real and imaginary parts of S-parameters.