Application of Different Methods to Quantify Uncertainty in Specific Absorption Rate Calculation Using a CAD-Based Mobile Phone Model

A Statistical Parsimony Method for Uncertainty Quantification of FDTD Computation Based on the PCA and Ridge Regression

“…Various sparse techniques can be utilised to alleviate the curse of dimensionality. However, the computational cost is [52] AND [20], WHERE P-PCA-RR REFERS TO THE PARSIMONIOUS PCA-RR BASED METHOD WHICH ONLY USES THE MINIMUM NUMBER OF REQUIRED FDTD SIMULATIONS, AND HA-PCA-RR INDICATES THE HIGH-ACCURACY PCA-RR BASED METHOD WHOSE MINIMUM ε l IN 18 still non-negligible when a system has a considerable number of input parameters. This paper proposed a versatile UQ technique in order to overcome this problem.…”

“…The transformed input parameters are then utilised to construct the sparse PC expansion based on the hyperbolic scheme. As a consequence, only a quarter of the PC bases are required, in comparison to the number utilised for NIPC expansion via general truncation [20]. • Uncertainty quantification: We adopt the ridge regression (RR) method [46], [47] to estimate the coefficients of PC bases.…”

“…We define the PCA-RR based method with the minimum ε l in (18) less than 0.1 as the high-accuracy PCA-RR (HA-PCA-RR) based method. Table II presents the comparisons of the proposed method with other advanced UQ techniques including the NIPC method, the works in [52] and [20], where the P-PCA-RR based method refers to the parsimonious PCA-RR based method which only uses the minimum number of required FDTD simulations. [52] combines PCA with sparse grid interpolation for UQ and the relative tolerance is set to 0.2 in our scenario.…”

“…[52] combines PCA with sparse grid interpolation for UQ and the relative tolerance is set to 0.2 in our scenario. [20] proposes a combined uncertainty based UQ method. It is worth noting that the accuracy of σ in HA-PCA-RR based method outperforms other UQ methods while it demands the second least number of FDTD simulations.…”

“…In the projection method, the coefficients in NIPC can be estimated by the quadrature rule or MCM, while the regression method is based on the least square minimisation of the error between the system output and its approximation. However, both methods are prone to the curse of dimensionality [20], whereby the number of required simulations grows substantially as the number of input parameters increases.…”