A Case for Hypergraphs to Model and Map SNNs on Neuromorphic Hardware
Marco Ronzani, Cristina Silvano
TL;DR
This work addresses the challenge of mapping large-scale spiking neural networks to neuromorphic hardware by reframing SNNs as directed hypergraphs to capture multicast spike replication. It introduces two guiding mapping affinities—synaptic reuse (second-order) and connections locality (first-order)—and develops a suite of hypergraph-based partitioning and placement algorithms, including a novel hyperedge overlap-based partitioning method and spectral placement with force-directed refinement. The approach yields mappings with superior energy-latency performance and scalability across layered, recurrent, and bio-plausible SNNs, corroborated by experiments on small and large neuromorphic hardware analogs. By formalizing the problem and validating the proposed methods, the work demonstrates that hypergraph-based mapping can enable brain-scale SNN deployments on scalable neuromorphic platforms, with open-source benchmarks to foster further research.
Abstract
Executing Spiking Neural Networks (SNNs) on neuromorphic hardware poses the problem of mapping neurons to cores. SNNs operate by propagating spikes between neurons that form a graph through synapses. Neuromorphic hardware mimics them through a network-on-chip, transmitting spikes, and a mesh of cores, each managing several neurons. Its operational cost is tied to spike movement and active cores. A mapping comprises two tasks: partitioning the SNN's graph to fit inside cores and placement of each partition on the hardware mesh. Both are NP-hard problems, and as SNNs and hardware scale towards billions of neurons, they become increasingly difficult to tackle effectively. In this work, we propose to raise the abstraction of SNNs from graphs to hypergraphs, redesigning mapping techniques accordingly. The resulting model faithfully captures the replication of spikes inside cores by exposing the notion of hyperedge co-membership between neurons. We further show that the overlap and locality of hyperedges strongly correlate with high-quality mappings, making these properties instrumental in devising mapping algorithms. By exploiting them directly, grouping neurons through shared hyperedges, communication traffic and hardware resource usage can be reduced be yond what just contracting individual connections attains. To substantiate this insight, we consider several partitioning and placement algorithms, some newly devised, others adapted from literature, and compare them over progressively larger and bio-plausible SNNs. Our results show that hypergraph based techniques can achieve better mappings than the state-of-the-art at several execution time regimes. Based on these observations, we identify a promising selection of algorithms to achieve effective mappings at any scale.