Reversible Gating Architecture for Rare Failure Detection of Analog and Mixed-Signal Circuits
TimeWednesday, December 8th3:30pm - 3:50pm PST
Manufacturing Test and Reliability
DescriptionDue to the expensive simulation and manufacturing cost in analog and mixed-signal (AMS) circuit design, AMS failure detection in high-dimensional variational space is demanding with limited samples. In this work, we propose a reversible gating architecture to identify essential features from datasets and reduce feature dimension for the failure detection. Through reversible residual networks (RevNets) for their restoration ability from output to input without loss of information, the proposed gating architecture is incorporated into Bayesian optimization (BO) framework to reduce the dimensionality of BO embedding important features. Furthermore, two restoration schemes in the RevNet are explored for efficient dimension reduction.