A Jackknife Variance-based Stopping Criterion for Adaptive Verification of Integrated Circuits

Adaptive verification appears to be an efficient solution for the time}-consuming simulations of complex integrated circuits with multiple inputs and outputs. Compared to the traditional (one-shot) experimental design, in which all samples are selected andsimulated beforehand, adaptive sampling is an iterative method in which the information acquiredfrom previously evaluated samples are used to better understand and interpret the behavior of the system. However, if no information is available upfront, it can be very difficult to determine the total number of experiment runs (simulationslmeasurements) that will be needed. This paper proposes a stopping criterion for planning adaptive experiments using the concept of jackknife variance and helps to efficiently characterize the behavior of complex systems with a number of experiment runs as low as possible. The approach was evaluated on synthetic testfunctions and an electronic system. The stopping criterion shows substantial improvements in the number of experiment runs while maintaining high resolution in the region of interest.