A new modeling technique based on the ANN and DOE for interconnects

This paper mainly introduces a new method for building interconnect macro-models based on the multilayer feed-forward artificial neural network (ANN) and design of experiment (DOE). The test examples show that making use of orthogonal tables to arrange the selection of training data can obviously improve the efficiency for building macro-models of interconnect networks in comparison with training from random data. The method can also be used to build macro-models of other devices or modules of VLSI or RF/MMICs.