Zoom Link: https://hku.zoom.us/j/97430126742?pwd=ou6CUPNMjhlrmRbwUKRa8aTHi6PjYX.1
Meeting ID: 974 3012 6742
Password: 967270
Abstract
The human brain operates as a sophisticated spiking neural network (SNN), capable of learning multimodal signals in a zero-shot manner by leveraging prior knowledge. Impressively, it accomplishes this with minimal energy consumption, relying on event-driven signals that travel through its intricate structure. However, replicating the brain’s functionality in efficient neuromorphic hardware poses significant challenges in both hardware and software. Moreover, training these algorithms demands extensive resources, and effective security measures remain insufficient.
Benefiting from the RRAM array inherit stochasticity, we demonstrated an efficient analogue-digital system that can handle multi-modal spiking signals and possess zero-shot learning capability like a human. This reservoir accelerated system enables significant lower training overheads while maintaining comparable baseline utility. Since emerging brain-inspired computing raises security concerns, we also share new methodologies insights onto these neuromorphic systems that can secure non-volatile CIM-based parameters without sacrificing latency and energy efficiency.
This presentation will delve into the development of a secure and efficient brain-inspired in-memory computing system, achieved through the integrated co-design of algorithms, circuits, and devices.
Speaker
Mr. WONG Edwin Kwun Hang
Department of Electrical and Electronic Engineering
The University of Hong Kong
Speaker’s Biography
Mr. Edwin Kwun Hang Wong received the B.Eng. (EE) degree from The University of Hong Kong in 2023. He is currently working towards MPhil degree with The University of Hong Kong. His research interests include AI Security, Brain-inspired computing, and RRAM-based accelerator.
All are welcome!