Spiking Neural Networks (SNNs) have emerged as an alternative to deep learning especially for edge computing due to their huge energy efficiency benefits on neuromorphic hardware. In this presentation, I will discuss the roadmap of current activities in the SNN algorithm and hardware design space. Particularly, I will describe our group’s recent works towards enabling and democratizing spike-based machine intelligence design, simulation, and evaluation across different applications. I will talk about the importance of temporal dimension in SNNs which unlock unique behavior such as, robustness and bring in huge benefits in terms of latency, energy, and accuracy in different applications like video segmentation, human activity recognition, event sensing among others. Then, I will delve into the hardware perspective of SNNs and the prospects around memory and sparsity management for accelerating SNNs on general purpose platforms. I will highlight some techniques such as, input-aware dynamic temporal exit and membrane-potential sharing across time. Finally, I will discuss a future landscape for hardware-software co-design for spike-based on-device intelligence.

Location: SEC 1.307 and Zoom (Password: 599436)