Multi-Corner Parametric Yield Estimation via Bayesian Inference on Bernoulli Distribution with Conjugate Prior

Automatic Design of a Broadband Directional Coupler via Bayesian Optimization

“…Parametric yield estimation [120]- [123] is closely related to circuit performance modeling as shown in Eq. (21).…”

“…However, this Gaussian prior makes the posterior intractable, and a Laplacian approximation is adopted to do posterior inference [120]. Alternatively, [123] exploits a conjugate prior, so the posterior has a closed form and the computation is straightforward. Both these works demonstrate better accuracy or around 2× reduced sampling cost when compared with Monte Carlo.…”

“…Rare Failure Rate w/ P-L-P (MAP or MLE) [120], [121], [123], [125] [126]- [132] w/o P-L-P (Sampling method) × [124], [133]- [156] * P-L-P is short for prior-likelihood-posterior. 'w/ P-L-P' implies that the Bayesian framework (i.e., maximum-a-posteriori) or at least likelihood function and maximizing likelihood estimation is used.…”

“…In [120] and [123], the authors consider the multi-corner yield estimation problem. Namely, a set of yield values {P 1 s, , P 2 s, , • • • , P K s, } are desired to be estimated, where K represents the number of process corners.…”