Development of Improved Predictor for Expedited Training of Polynomial Chaos Metamodels of Multi-Walled Carbon Nanotube Interconnects

This paper presents an improved predictor-corrector algorithm to efficiently train polynomial chaos (PC) metamodels for the variability analysis of multiwalled carbon nanotube (MWCNT) interconnect networks. The salient feature of the proposed algorithm is the development of a more accurate predictor that can accelerate the convergence of conventional predictor-corrector algorithms. Therefore, the proposed predictor-corrector algorithm offers much greater speedup than conventional predictor-corrector algorithms when training PC metamodels.