A combination of FDTD and least-squares support vector machines for analysis of microwave integrated circuits

This paper presents a new combination of the finite-difference time-domain (FDTD) method and the least-squares support vector machines (LS-SVM) technique. The LS-SVM is a statistical-learning method which has a self-contained basis of statistical-learning theory and excellent learning performance. A short segment of an FDTD record is used to train the LS-SVM predictor in order to obtain an accurate future realization. Numerical simulations for two typical microwave filters demonstrate that the LS-SVM method can achieve good forecasting accuracy and the efficiency of the FDTD method can be improved by up to 70%. © 2005 Wiley Periodicals, Inc. Microwave Opt Technol Lett 44: 296–299, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mop.20615