Effects of Random Number Generations on Intelligent Semiconductor Device Model Parameter Extraction

Abstract

In this work, we experimentally compare the effect of random number generations on the performance of semiconductor device model parameter extraction. Based upon the genetic algorithm, the neural network and the Levenberg-Marquardt method, the prototype of parameter extraction has been developed in our earlier work. Property of the evolutionary technique is further advanced by implementing eight different random number generation schemes, where convergent behavior is compared. For both extraction cases of single and multiple nanoscale devices, the chaotic random number generator possesses superior convergence behavior than other random number generation methods. It generates the random numbers with better distribution which keeps the high diversity of the extraction system, thus the best performance of the convergence score is reached.