Technology that analyzes the brain's neural network consolidation is drawing attention as a core foundation of brain-computer interface (BCI) technologies, such as controlling artificial limbs and augmenting human intelligence. To perform such analysis more precisely, it is most important to interpret quickly and accurately the complex signals generated by countless neurons in the brain.
A team led by Senior Researcher Park Jong-gil of the Korea Institute of Science and Technology (KIST) Semiconductor Engineering Research Division presented a new approach that mimics the brain's learning principles. The team said on the 29th that it has engineered the principle of "spike-timing-dependent plasticity (STDP)," in which the brain adjusts consolidation strength between neurons according to the order in which signals occur. Through this, the team developed technology that can learn the consolidation relationships of neural networks in real time without storing all neuronal activity.
Existing technology stores neuronal activity data for a long time and then calculates consolidation relationships between neurons using statistical methods. As the size of the neural network grows, this method incurs massive computational loads and time delays, making real-time analysis virtually impossible in environments where countless signals occur simultaneously, like the brain.
The team devised a new learning architecture that can drastically reduce the large memory required when implementing "spike-timing-dependent plasticity (STDP)" in hardware. As a result, the hardware-based "on-chip learning neuromorphic system" developed this time achieved processing speeds up to 20,000 times faster while maintaining analysis accuracy similar to existing technologies.
"Neuromorphic" technology is a next-generation artificial intelligence (AI) semiconductor that mimics the brain's neural network structure and learning methods to simulate human cognitive abilities, and is a strategic field into which major advanced countries, including the United States and Europe, are investing heavily to secure technological supremacy. However, commercialization has been difficult due to a lack of concrete application fields that can be as useful as the brain, namely "killer applications." In this context, the "real-time brain neural network consolidation structure analysis" technology proposed by the KIST team is a case that proves the practical applicability of neuromorphic technology.
Senior Researcher Park Jong-gil said, "This achievement will be an important turning point for Neuromorphic Computing to evolve into a powerful tool for solving real-world problems," adding, "Because the hardware architecture is simple and easy to scale, in the future it could be applied to advanced AI fields such as controlling devices with thought alone or copying specific brain functions, as well as analyzing in real time complex sensor signals where time order and cause-and-effect relationships are critical, such as in Autonomous Driving and satellite communications."
The results of this study were published in July in the international journal "Institute of Electrical and Electronics Engineers (IEEE) Transactions on Neural Systems and Rehabilitation Engineering (TNSRE)."
References
IEEE Transactions on Neural Systems and Rehabilitation Engineering (2025), DOI: https://doi.org/10.1109/TNSRE.2025.3583057