美国伊利诺伊大学香槟分校 Naresh R. Shanbhag教授到访微纳电子学系
2019年5月14日，来自美国伊利诺伊大学香槟分校的Naresh R. Shanbhag教授应邀来访微纳电子学系，并在微纳电子大厦103报告厅做了题为《Bringing Artificial Intelligence to the Edge2 via Shannon-inspired Statistical Computing》的学术报告。微纳电子学系多位老师以及研究生、本科生参加了报告会，并与报告人进行了热烈的讨论。
Bringing Artificial Intelligence to the Edge2 via Shannon-inspired Statistical Computing
Much of AI today is deployed in the Cloud primarily due to the high complexity of machine learning algorithms. Realizing inference functionality on sensory Edge devices requires one to find ways to operate at the other edge, i.e., at the limits of energy efficiency, latency, and accuracy, in nanoscale semiconductor technologies. This talk will describe a Shannon-inspired model of computing (Proceedings of the IEEE, January 2019) to accomplish this objective. This framework comprises low signal-to-noise ratio (SNR) circuit fabrics (the channel) with engineered error statistics, coupled with efficient techniques to compensate for computational errors (encoder and decoder). A low SNR circuit fabric referred to as deep in-memory architecture (DIMA) will be described. DIMA breaches the long-standing “memory wall” in von Neumann architectures by embedding analog computations in the periphery of the memory array (see https://spectrum.ieee.org/computing/hardware/to-speed-up-ai-mix-memory-and-processing) thereby achieving >100X energy-delay-product gains in laboratory prototypes over custom digital architectures implementing the same inference function. Other examples of Shannon-inspired design methods include designing deep learning systems in fixed-point, energy efficient subthreshold ECG classifier ICs, and STT-RAM based all-spin logic competitive with CMOS. The strong systems-to-devices connection inherent in Shannon-inspired statistical computing creates an opportunity for researchers in machine learning, computer architecture, integrated circuits, and nanoscale devices, to collaborate in designing intelligent machines of the future. The talk will conclude with a discussion of future directions.