The neuromorphic memory device is a simulation of neurons and synapses
Researchers have developed a nano-sized memory device that simulates synapses and neurons simultaneously within a single unit cell. This is another step towards the completion of the goal for neuromorphic computing, which is designed to mimic the human mind with semiconductor devices.
Neuromorphic computing is a way to achieve artificial intelligence by replicating the synapses and neurons that make up the brain. Neuromorphic devices are based on the cognitive functions that the brain can provide but cannot be replicated by current computers. Current Complementary Metal-Oxide Semiconductor-based neuromorphic devices connect artificial synapses and neurons without synergistic interaction, but the simultaneous implementation of synapses and neurons remains a challenge. In order to address these issues, Professor Keon Jae Le from the Department of Materials Science and Engineering led a team of researchers who implemented the biological working mechanism of humans through the introduction of neuron-synapse interaction in a single cell of memory, instead of the conventional approach of connecting artificial neuronal and synapsic devices electrically.
The artificial synaptic devices studied previously were often used to speed up parallel computations. However, they differed from the operating mechanisms of the brain. The team of researchers implemented synergistic interaction between synapses and neurons in the neuromorphic device, simulating mechanisms from the biological neural networks. The developed neuromorphic memory device is able to replace complex CMOS circuits, allowing for high scalability, and cost-efficiency.