from the training sample set. For comparison, conventional threelayer MLP models are also developed with the same sets of training and testing data. Figure 3 shows the responses given by the FDTD simulation, the MLP models, and the NNKBN models. It is obvious that the NNKBN models have much better results than those given by the MLP models.The extrapolation properties of the NNKBN and MLP models are explored and compared. The input data of w beyond the range of training sample set by 50% are selected to test the two kinds of models. With the use of the FDTD simulation data as desired outputs, the average absolute extrapolating errors (AAEE) are given in Table 2. From Table 2, it is found that the NNKBN models have great improvement in extrapolation.
CONCLUSIONCompared with the conventional MLP model, the NNKBN model requires fewer neurons in the hidden layer, and has better accuracy with less training samples and higher reliability of extrapolation.The NNKBN model preserves the accuracy of the full-wave FDTD simulations yet simplifies their CPU requirement. Furthermore, it maintains reliable finite extrapolation capability by incorporating the analytic prior knowledge. The potential power of the NNKBN model is useful for the interactive CAD of the stripline interconnect design in the HSDICs.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.