Amith Singhee Carnegie Mellon University

More Monte Carlo!

For highly replicated circuits such as SRAMs and flip flops, a rare statistical event for one circuit may induce a not-so-rare system failure. Existing techniques perform poorly when tasked to generate both efficient sampling and sound statistics for these rare events. Regular Monte Carlo can require millions and millions of simulations to extract statistics of such rare failures. Other analytical methods make unreasonable assumptions that are not valid for these rare events.

I will present Statistical Blockade: a novel Monte Carlo technique that allows us to efficiently filter (to block) unwanted samples insufficiently rare in the tail distributions we seek. The method synthesizes ideas from data mining and Extreme Value Theory, and shows speedups of 10X -100X over standard Monte Carlo.

Bio:

Amith Singhee is a PhD student in ECE. He got his B.Tech. from the Indian Institute of Technology, Kharagpur, in 2000, and MS from CMU in 2002.  From 2002-04 he worked at Neolinear, and Cadence Design Systems, as a lead researcher for the NeoCircuit analog synthesis tool. His interests lie in modeling, simulating and optimizing circuits, especially in the presence of random manufacturing variations.