Chapter 6
Evaluation Results
To verify the accuracy of the proposed methods, we simulated variations in a
number of MCNC benchmark circuits. Then, we used ^ɪ-regularization, l∙2-
minimization and TUSC (see Section 3.4) to estimate the variations. The simula-
tion result shows that using ^ɪ-regularization and TUSC improve the estimations
dramatically.
6.1 Simulations setup
• The variation model: As it is explained in Section 2.2.1, we have used
multivariate Gaussian distribution to model the spatial correlation in the
variations. The model well agrees with the measurement data and is also
used by other researchers [22,32,47,69].
65