Energy-efficient artificial intelligence hardware technology via a brain-inspired storage system


Researchers demonstrate a neuromodulation-inspired storage system for energy-efficient learning of a spiked neural network using a self-rectifying memristor array

Image: A diagram illustrating localized brain activity (ac) and hybrid hardware and software neural network (de) configuration using a self-rectifying memristor (fg) network.

Researchers have proposed a new system inspired by neuromodulation of the brain, called a “storage system”, which requires less energy consumption. The research group led by Professor Kyung Min Kim from the Department of Materials Science and Engineering has developed a technology capable of efficiently handling mathematical operations for artificial intelligence by mimicking continuous changes in neural network topology depending on the situation. The human brain changes its neural topology in real time, learning to store or recall memories as needed. The research group presented a new method of learning artificial intelligence that directly implements these configurations of neural coordination circuits.

Research on artificial intelligence is becoming very active, and the development of artificial intelligence-based electronic devices and product launches are accelerating, especially in the era of the fourth industrial revolution. To implement artificial intelligence in electronic devices, the development of custom hardware must also be supported. However, most electronic devices for artificial intelligence require high power consumption and highly integrated memory arrays for large-scale tasks. It has been difficult to solve these power consumption and integration limitations, and efforts have been made to find out how the human brain solves the problems.

To prove the effectiveness of the developed technology, the research group created an artificial neural network hardware equipped with a self-rectifying synaptic network and an algorithm called “storage system” which was developed to perform the artificial intelligence learning. As a result, he was able to reduce energy by 37% in the storage system without any degradation in accuracy. This result proves that it is possible to emulate neuromodulation in humans.

Professor Kim said: “In this study, we implemented the human brain learning method with only a simple circuit composition and through this we were able to reduce the energy needed by almost 40%. “

This neuromodulation-inspired storage system that mimics neural activity in the brain is compatible with existing electronic devices and commercial semiconductor hardware. It is expected to be used in the design of next-generation semiconductor chips for artificial intelligence.

This study was published in Advanced Functional Materials in March 2022 and supported by KAIST, National Research Foundation of Korea, National NanoFab Center and SK Hynix.


Woon Hyung Cheong, Jae Bum Jeon†, Jae Hyun In, Geunyoung Kim, Hanchan Song, Janho An, Juseong Park, Young Seok Kim, Cheol Seong Hwang and Kyung Min Kim (2022)

“Demonstration of a neuromodulation-inspired storage system for energy-efficient learning of a spiked neural network using a self-rectifying memristor network”, Advanced Functional

Documents March 31, 2022 (DOI: 10.1002/adfm.202200337)

/Public release. This material from the original organization/authors may be ad hoc in nature, edited for clarity, style and length. The views and opinions expressed are those of the author or authors.View Full here.

About Author

Comments are closed.