Artificial intelligence (AI) and the Internet of Things (IoT) are revolutionizing the way we live. Neural networks form the vast majority of algorithms used by current AI architectures. But the current generation of neural networks also have their drawbacks: they are notoriously energy-hungry and need a lot of data to be trained effectively. Those two facts restrict their deployment at the edge of netwoks as IoT devices need to remain energy-efficient and embedded processing units based on the classical von-Neumann architecture used in current IoT solutions suffer from memory bottlenecks. This PhD research is tackling those two constraints of classical neural networks by studying so-called spiking neural networks (SNN) which are neuromorphic models of computation inspired by the inner-working of biological neural systems. SNNs promise to be energy-efficient, to learn and infere at faster rates and to allow on-line learning which classical on-chip neural networks can simply not do. This PhD research will build SNN algorithms and a neuromorphic chip dedicated to radar-assisted drone navigation which has never been done before. The results are expected to be very close to market-ready.
7/6/2024 10:00 - 12:00
Aula Wolfspoort
