Effective and Fast Learning for Neuromorphic Hardware

Spiking neural networks (SNNs) are the basis for many energy-efficient neuromorphic hardware systems. While the standard neuron model for spike-based computation on such systems has long been the leaky integrate-and-fire (LIF) neuron, a computationally light augmentation with an adaptation mechanism has recently been shown to exhibit superior performance on spatio-temporal processing tasks. I will first discuss the dynamical, computational, and learning properties of adaptive LIF neurons and networks thereof. I will show that challenges related to stability of this model can be effectively, allowing to improve over state-of-the-art performances on common event-based benchmark datasets. In the second part of my talk, I will discuss another important topic in the field. In contrast to current neuromorphic systems, the brain exhibits the remarkable ability to learn from a limited number of examples and to quickly adapt to new tasks. These capabilities are arguably essential for intelligent neuromorphic systems in real-world scenarios. I will present methods that allow few-shot learning in neuromorphic hardware and demonstrate their applicability to memristor-based in-memory computing hardware.



The talk will be broadcast in the following zoom link: https://zoom.us/j/98286706234?pwd=bm1JUFVYcTJkaVl1VU55L0FiWDRIUT09