Advancing SNNs on low-power FPGAs

This presentation investigates the challenges of combining the advantages of transformer-inspired attention mechanism and event-driven sparse computation in lightweight Spiking Neural Networks for low-power real-time inference. To leverage these advantages on the wearable domain, a compact hardware architecture tailored for spiking transformers deployment on resource-constrained FPGAs was developed. Compared to the alternatives in the literature, limited to high-end FPGA platforms, we focus on resource minimization, demonstrating suitability for real-time execution in biological signal processing on the Lattice iCE40UP5K.