2018 ISIT Tutorial
Vail, CO - 6/17/18
Nanoscale Information Processing
Lav R. Varshney and Naresh R. Shanbhag
This tutorial highlight a potential role for information theorists in ensuring the continuation of Moore’s Law, which has been the driving force behind the exponential growth of the semiconductor industry for the past five decades.
Today, energy efficiency and reliability challenges at the end of the CMOS roadmap and in beyond-CMOS nanoscale technologies threaten the continuation of Moore’s Law. With low‐level device and circuit solutions running out of steam, there is growing interest in bringing inspiration from Shannon information theory. The idea is to have a statistical information processing framework that treats the problem of energy‐constrained computing on unreliable fabrics akin to information transfer over a noisy channel and seeks to transform computing from its deterministic roots to statistical information processing. Key elements of this framework are the use of information theory to determine fundamental limits on information processing capacity given unreliable nanoscale fabrics, and further to use statistical detection, estimation, and machine learning principles to design robust subsystems and programming models for computation, communication, and storage.
The tutorial will discuss stochastic models of nanoscale devices; provide examples to demonstrate current circuit and system design principles; present important open questions in statistical information processing; and suggest several information-theoretic techniques that may be relevant. The tutorial is largely based on work completed in the five-year ($30M) Systems on Nanoscale Information fabriCs (SONIC) center, one of the six SRC STARnet Centers, sponsored by MARCO and DARPA .