A Complete Pipeline for deploying SNNs with Synaptic Delays on Loihi 2
Balázs Mészáros, James C. Knight, Jonathan Timcheck, Thomas Nowotny
TL;DR
This work tackles edge-friendly keyword recognition by delivering a complete end-to-end pipeline that trains SNNs with learnable synaptic delays using EventProp on GPUs and deploys them to the Loihi 2 neuromorphic chip. The method combines mlGeNN for GPU-based training, a NetX exporter for Loihi deployment, and a delay-enabled learning framework, demonstrating substantial training efficiency improvements and notable accuracy gains when delays are used. Key findings show that delays yield significant accuracy boosts on Loihi, especially for feedforward architectures, while Loihi 2 achieves low-energy, low-latency inference with only modest overhead from delays; in benchmarks, neuromorphic deployment offers sizable advantages over conventional embedded GPUs. This work advances practical neuromorphic edge computing for speech recognition by enabling end-to-end learning with delays on Loihi 2, informing design choices around delay limits, batch processing, and quantization for real-time, energy-efficient keyword detection.
Abstract
Spiking Neural Networks are attracting increased attention as a more energy-efficient alternative to traditional Artificial Neural Networks for edge computing. Neuromorphic computing can significantly reduce energy requirements. Here, we present a complete pipeline: efficient event-based training of SNNs with synaptic delays on GPUs and deployment on Intel's Loihi 2 neuromorphic chip. We evaluate our approach on keyword recognition tasks using the Spiking Heidelberg Digits and Spiking Speech Commands datasets, demonstrating that our algorithm can enhance classification accuracy compared to architectures without delays. Our benchmarking indicates almost no accuracy loss between GPU and Loihi 2 implementations, while classification on Loihi 2 is up to 18x faster and uses 250x less energy than on an NVIDIA Jetson Orin Nano.