Beschrijving
Undoubtedly, we are still far from reproducing the exceptional efficiency, speed, and intelligence of natural brains with our current computing systems. Modern computing systems are increasingly facing fundamental limitations, such as the memory bottleneck and the end of Dennard scaling, where gains in performance per unit of energy consumption are drastically diminishing. This has led researchers around the globe to explore new computational designs inspired by the brain. Neuromorphic computing is a leading solution in developing energy-efficient hardware capable of handling real-time signal processing and facilitating various edge artificial intelligence applications. This approach uses event-driven, massively parallel spiking neural networks to replicate brain-like computation at the hardware level.In my lecture, I will delve into the concept of spiking neural networks, explaining modern training techniques and their hardware implementations that emphasize event-driven sensing and computing. Neuromorphic systems result in embedded ultra-low-power implementations that support diverse applications in the smart sensing domains. I will conclude by demonstrating several prototype devices that adhere to the rigorous energy and cost-saving requirements of the Internet of Things (IoT) and support biomedical signal processing and radar sensing applications. These devices pave the way for the application of spiking neural networks in embedded sensing and computing devices.
| Periode | 31 jul. 2024 |
|---|---|
| Evenementstitel | Summer School of Information Engineering (SSIE): Machine Learning and its Applications: The Road Towards 6G Networks |
| Evenementstype | Workshop |
| Locatie | Bressanone, ItaliëToon op kaart |
| Mate van erkenning | Internationaal |