Military researchers to brief industry in May on ScAN artificial intelligence (AI) analog neural networks
ARLINGTON, Va. – U.S. military researchers will brief industry next month on an upcoming project to develop large-scale analog neural networks computing technology that could interface directly with the analog outputs of conventional sensors.
Officials of the U.S. Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., will host virtual industry briefings from 1 to 4:30 p.m. on 15 May 2024 on the Scalable Analog Neural-networks (ScAN) project.
Analog computing partially has been introduced into otherwise digital neural network systems to increase their power efficiency, yet these in-memory compute architectures are reliable only at small scales, which require the digital storage of intermediate results, which greatly reduces potential efficiency gains.
Neural network technology is part of artificial intelligence (AI), and is a machine learning program able to make decisions similar to those of the human brain. Neural networks use processes that mimic the way biological neurons work together in the human brain to identify phenomena, weigh options, and arrive at conclusions.
The ScAN program seeks to develop new analog neural networks that could interface directly with the analog outputs of conventional sensors and demonstrate a three-order-of-magnitude power reduction over existing solutions.
ScAN will develop new analog neural network processing algorithms and architectures that achieve state-of-the-art inferencing accuracy; robustness to process, voltage, and temperature variations; and scalability.
The ScAN project will be a 54-month, two-phase program to develop robust, accurate, and power-efficient analog neural network circuits; and find ways to scale-up analog neural networks to large-scale systems. A formal solicitation for the ScAN program is expected in May.
DARPA researchers first want to techniques to develop address the technical challenges to realize robust, accurate, and power-efficient analog neural network circuits at an intermediate-scale with a focus on scalable methods.
Related: The coming of age of artificial intelligence
Second, researchers want to develop ways to scale analog neural networks to large-scale systems that require only device-independent training while maintaining power efficiency, accuracy, and robustness.
Companies interested in attending the ScAN virtual briefings should register no later than 8 May 2024 online at https://web.cvent.com/event/ed21f6f8-9cb8-4d28-96f2-ff39489e9d75/summary. The virtual briefings will be via ZoomGov.
Email questions or concerns to Bryan Jacobs, the DARPA ScAN program manager, at [email protected]. More information is online at https://sam.gov/opp/dab3fc13713f456d981a2a9bd516e580/view.
John Keller | Editor-in-Chief
John Keller is the Editor-in-Chief, Military & Aerospace Electronics Magazine--provides extensive coverage and analysis of enabling electronics and optoelectronic technologies in military, space and commercial aviation applications. John has been a member of the Military & Aerospace Electronics staff since 1989 and chief editor since 1995.
