Implementing Neuromorphic Computing: Connecting Artificial Neurons and Synapses in Hardware


The development of artificial intelligence (AI) technologies has led to an increased focus on the creation of next-generation AI semiconductors that can handle large amounts of data efficiently. One area of interest is neuromorphic computing, which aims to mimic the brain’s structure and function. While individual devices that imitate neurons and synapses have been successfully developed, there is still a need for research on integrating these devices into a cohesive system to optimize their performance.

To address this challenge, a team led by Dr. Joon Young Kwak from the Korea Institute of Science and Technology (KIST) has implemented an integrated element technology for artificial neuromorphic devices. Their approach involves connecting neurons and synapses in a modular fashion, similar to stacking Lego blocks, to construct large-scale artificial neural network hardware. The team utilized hBN, a two-dimensional material known for its high integration capabilities and ultra-low power consumption, to fabricate vertically-stacked memristor devices that exhibit characteristics similar to biological neurons and synapses.

Unlike conventional artificial neural network devices that rely on complex structures and multiple devices, the team designed their artificial neuron and synaptic devices with the same material and structure. This simplifies the fabrication process and allows for easy scalability of the network, making it more practical for large-scale implementation. The researchers successfully demonstrated the implementation of the neuron-synapse-neuron structure, which serves as the basic building block of an artificial neural network. This structure enables spike signal-based information transmission, mirroring the operation of the human brain.

The team also conducted experiments to adjust the modulation of spike signal information between two neurons based on the synaptic weights of the artificial synaptic device. This demonstrates the potential of using hBN-based emerging devices for low-power, large-scale AI hardware systems. Beyond its applications in AI, this technology can find utility in various real-life scenarios such as smart cities, healthcare, next-generation communications, weather forecasting, and autonomous vehicles. By significantly reducing energy consumption beyond the limits of existing silicon CMOS-based devices, it can also contribute to mitigating environmental concerns such as carbon emissions.

The development of integrated artificial neuromorphic devices marks a significant advancement in the field of AI hardware. By leveraging hBN’s unique properties, researchers have successfully connected artificial neurons and synapses, paving the way for the implementation of large-scale artificial neural network hardware. This breakthrough has the potential to revolutionize various industries and enhance the efficiency of data processing tasks, while also addressing environmental challenges associated with energy consumption. As further research and development are conducted, this technology could become a key contributor to the advancement of AI and its applications in the modern world.

1. Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it