Researchers at the Korea Advanced Institute of Technology (KAIST) have made brain-inspired neuromorphic equipment by combining single-transistor neurons and synapses. The device used standard silicon complementary metal oxide semiconductor (CMOS) technology. It will reduce the cost of microcircuits and simplify the manufacturing procedure.
Scientists have created neurons and synapses from a single transistor for highly scalable neuromorphic equipment. They also showed that this model is capable of recognizing text and facial images. Research on this appeared in the journal Science Advances.
Neuromorphic equipment works like an AI-based model, but consumes minuscule amounts of energy, mimicking the human brain. Neuromorphic equipment requires a neuron that generates power when a certain signal is integrated, and a synapse that remembers the connection between two neurons, just like the human brain. But since neurons and synapses built on digital or analog circuits take up a lot of space, it has a limit in efficiency and a limitation in terms of hardware cost.
To solve this problem, the research team mimicked the behavior of biological neurons and synapses with a single transistor and combined them on an 8-inch plate. Manufactured neuromorphic transistors have the same structure as transistors for memory and logic, which are mass-produced.
This work could significantly reduce the cost of hardware by replacing neurons and synapses, which were based on complex digital and analog circuits, with a single transistor. “We have shown for the first time that neurons and synapses can be used with a single transistor,” the researchers noted.