top of page
Brain-inspired AI
The human brain operates far more efficiently than modern deep learning systems.
Our research focuses on understanding the learning mechanisms of the brain and leveraging these insights to develop novel AI frameworks.
SNN
Spiking Neural Networks (SNN)
We study SNNs that compute using discrete "spikes", emulating the temporal and energy-efficient processing of the brain. Our research explores how event-driven signals can process complex temporal data and how these can be integrated into high-performance architectures. By evolving SNNs into scalable frameworks, we aim to achieve high-intelligence systems that consume minimal power.
Local Learning
Local Learning
Because the brain lacks a global error-correction system, intelligent agents must be able to update knowledge through local interactions. We research "Non-BP" (Non-Backpropagation) methods where neurons adjust their connections based on immediate, localized signal changes. This approach enables real-time, on-device learning that is robust against changing environments without catastrophic forgetting.
Efficient Learning
Efficient Learning & Foundation Models
We ultimately seek to scale these brain-inspired principles into large-scale foundation models. By combining SNN dynamics with scalable learning frameworks, we build AI that is both highly intelligent and resource-sustainable. This research bridges the gap between neural science and practical, large-scale AI deployment for the future of technology.
bottom of page