Researchers have managed to successfully emulate the synergistic interactions between neurons and synapses in neuromorphic circuits, pushing the frontiers of brain-like artificial intelligence.
Researchers have invented a new nano-sized neuromorphic memory device that emulates the functions of neurons and synapses simultaneously in a unit cell, pushing the frontiers of neuromorphic computing.
Neuromorphic computing is a branch of computing that explores the replication of human brain functions with semiconductor technology to enable more powerful artificial intelligence capable of cognitive function that rivals that of humans.
Unfortunately, the current Complementary Metal-Oxide Semiconductor (CMOS)-based neuromorphic circuit technology is unable to emulate the concomitant functions of neurons and synapses, as it simply connects the neurons and synapses, without taking into account the synergistic interactions that should also concurrently take place.
In an attempt to resolve these problems, a team of researchers led by Professor Keon Jae Lee from the Department of Materials Science and Engineering at the Korea Institute of Science and Technology developed a single synthetic memory cell that provides a more accurate technological implementation of the neuron-synapse interactions in humans, and the results were published in Nature Communications in May 2022.
The artificial synaptic devices that utilise the current CMOS technology accelerate parallel computations, just like computer graphics cards, but far from the neuronal signalling pathways in humans. In contrast, the new technology developed by the research team is capable of synergistic interactions between the artificial neurons and synapses in the neuromorphic cell, allowing for a closer resemblance of the devices to the mechanisms in the biological neural network. A further added advantage of this new technology is its scalability and cost-efficiency as each small device may replace entire complex CMOS neuron circuits.
The research team’s device emulates the flexibility of the functions and structures of the neurons and synapses that can flexibly change according to external stimuli.
It comprises volatile and non-volatile thin-film memory devices that emulate the characteristics of neurons and synapses. The neuron is emulated by a volatile memory device that is made of a threshold switch, while the synapse is emulated by a phase-change memory. Both devices are developed without intermediate electrodes, allowing for a compact memory cell.
Professor Keon Jae Lee explained, “Neurons and synapses interact with each other to establish cognitive functions such as memory and learning, so simulating both is an essential element for brain-inspired artificial intelligence. The developed neuromorphic memory device also mimics the retraining effect that allows quick learning of the forgotten information by implementing a positive feedback effect between neurons and synapses.”
Source: Sung et al. (2022). Simultaneous emulation of synaptic and intrinsic plasticity using a memristive synapse.Nature Communications,13(1), 1-12.