Neuromorphic computing technology that mimics the human brain must overcome limitations of excessive power consumption. This limitation is characteristic of the von Neumann computing method. To implement a brain-information transmission method using a semiconductor device, a high-performance analog artificial synapse must be developed. This technique transmits signals between neurons when a spike signal is generated by a neuron.
However with traditional resistance-variable memories devices widely used for artificial synapses. As the filament grows, the electric field will increase, causing a feedback phenomena, which results in rapid filament growth. It is difficult to achieve plasticity and maintain analog (gradual resistance variation) filament types with conventional resistance-variable memory devices.
The Korea Institute of Science and Technology led by Dr. YeonJooJeong’s team at Center for Neuromorphic Engineering solved the chronic problems of analog synaptic characteristics and plasticity, information preservation and information preservation. He announced the development of an artificial synaptic semiconductor device capable of highly reliable neuromorphic computing.
The KIST research team fine-tuned the redox properties of active electrode ions to solve small synaptic plasticity issues hindering the performance of existing neuromorphic semiconductor devices. The synaptic device also contained transition metals that were doped to control the active electrode ions’ reduction probability. Engineers discovered that the highest reduction probability of ions was a crucial variable in the design of high-performance artificial Synaptic Devices.
The research team introduced a titanium transitional metal with a high probability of reducing ions into an artificial synaptic device. The synapse retains its analog characteristics and the device’s plasticity at the synapse. This is approximately five times the difference in high and low resistances. Furthermore, they developed a high-performance neuromorphic semiconductor that is approximately 50 times more efficient.
Additionally, due to the high alloy formation reaction exhibited by the doped titanium transition metal, the information retention increased up to 63 times compared with the existing artificial synaptic device. Furthermore, brain functions, including long-term potentiation and long-term depression, could be more precisely simulated.
The team implemented an artificial neural network learning pattern using the developed artificial synaptic device and attempted artificial intelligence image recognition learning. The team was able to reduce the error rate by over 60%, and the recognition accuracy of handwriting patterns (MNIST), by 69%. This improved artificial synaptic device was confirmed by the research team as feasible.
Dr. KIST’s Jeong stated that “This study significantly improved the synaptic range and motion as well as information preservation which were the greatest technical obstacles of synaptic mimics.” The device’s analog operation area for expressing the various connections strengths of the synapse has been increased in the artificial synapse design. This will improve the performance and efficiency of brain-based artificial intelligence computing. “
The research was published in
Neuromorphic memory device simulates synapses and neurons
Jaehyun Kang et al, Cluster-type analogue memristor by engineering redox dynamics for high-performance neuromorphic computing, Nature Communications (2022). DOI: 10.1038/s41467-022-31804-4
National Research Council of Science & Technology
Engineers create high-performance, high-reliability artificial synaptic silicon devices (2022, September 20).
Retrieved 20 September 2022
Copyright is reserved for this document. Except for fair dealings for private study or research purposes, there is no
Part may not be reproduced without written permission. This content is only for informational purposes.