Quantification of Uncontrolled Large Variations of Parameters on System Performance via Support Vector Machine Surrogates

This paper presents a preliminary application of the support vector machine regression to the modeling and the statistical assessment of system responses affected by large uncertainties. The support vector machine regression is applied to generate a robust and compact surrogate model of the system output from a limited set of randomly selected training samples. The system surrogate can be suitably adopted both for design optimization and for stochastic analysis. The feasibility, accuracy and robustness to noise of the surrogate model calculated via the support vector machine regression are investigated by considering a realistic printed circuit board interconnect structure