During the 1990s, Carver Mead and his colleagues combined basic neuroscience research with the elegant design of analog circuits in electronic technology. This pioneering work on neuromorphic electronic circuits inspired researchers in Germany and Switzerland to explore the possibility of reproducing the physics of real neural circuits using silicon physics.
The field of brain-simulating neuromorphic electronics demonstrates great potential not only for basic research, but also for the commercial operation of constantly operating modern computing and the Internet of Things applications.
IN Letters on Applied Physics Elizabeth Chikka from the University of Bielefeld and Giacomo Indiveri from the University of Zurich and ETH Zurich present their work to understand how neural processing systems in biology perform computations, as well as a recipe for reproducing these computational principles in analog / mixed signal / digital electronics and new materials.
One of the most characteristic computing features of neural networks is training, which is why Chicca and Indiveri are especially interested in reproducing the adaptive and plastic properties of real synapses. They used both standard complementary metal-oxide semiconductor (CMOS) electronic circuits and advanced nanoscale memory technologies, such as storage devices, to create intelligent systems this can be learned.
This work is important because it can lead to a better understanding of how to implement sophisticated signal processing using extremely low-power and compact devices.
Their main findings are that the obvious drawbacks of these low-power computing technologies, mainly due to low accuracy, high sensitivity to noise and high variability, they can actually be used to perform reliable and efficient calculations, just as the brain can use highly variable and noisy neurons to implement reliable behavior.
Researchers say it is amazing to see the field of memory technology, usually associated with the exact bit technology of high-density devices, which now view the brain of animals as a source of inspiration for understanding how to create adaptive and reliable neural processing systems. This is very consistent with the main research program that Mead and his colleagues followed more than 30 years ago.
“The electronic neural processing systems that we create are not designed to compete with powerful and accurate artificial intelligence systems that work on energy-intensive large computer clusters to process a natural language or to recognize and classify high-resolution images,” said Cicca.
On the contrary, their systems "offer promising solutions for those applications that require a compact and very low power (submilliwatt) real-time processing with short delays, ”said Indiveri.
He said that examples of such applications belong to the field of “extreme computing”, which require a small amount of artificial intelligence to extract information from live or streaming sensory signals, such as processing biosignals in wearable devices, brain-machine interfaces, and constant environmental monitoring. "
"The recipe for creating ideal hybrid memristive-CMOS neuromorphic computing systems" Letters on Applied Physics (2020). DOI: 10.1063 / 1.5142089
American Institute of Physics
A special mixture of microchips and storage devices created for processing systems that simulate the brain (2020, March 24)
restored March 24, 2020
This document is protected by copyright. Other than honest deals for private study or research, no
Part may be reproduced without written permission. Content is provided for informational purposes only.