Sparse Basis Pursuit On Automatic Nonlinear Circuit Modeling

In this paper, we propose a black-box nonlinear dynamic modeling algorithm that automatically selects essential basis functions to overcome the overfitting problem. Our automatic modeling algorithm, which is formulated as a convex optimization problem, guarantees model stability in transient simulation. Furthermore, we incorporate our algorithm with a sparsity induction mechanism, which improves model robustness and generalization capabilities, as shown in our example.