EvtSNN: Event-driven SNN simulator optimized by population and pre-filtering

Abstract

Recently, spiking neural networks (SNNs) have been widely studied by researchers due to their biological interpretability and potential application of low power consumption. However, the traditional clock-driven simulators have the problem that the accuracy is limited by the time-step and the lateral inhibition failure. To address this issue, we introduce EvtSNN (Event SNN), a faster SNN event-driven simulator inspired by EDHA (Event-Driven High Accuracy). Two innovations are proposed to accelerate the calculation of event-driven neurons. Firstly, the intermediate results can be reused in population computing without repeated calculations. Secondly, unnecessary peak calculations will be skipped according to a condition. In the MNIST classification task, EvtSNN took 56 s to complete one epoch of unsupervised training and achieved 89.56% accuracy, while EDHA takes 642 s. In the benchmark experiments, the simulation speed of EvtSNN is 2.9–14.0 times that of EDHA under different network scales.

SHIP: a computational framework for simulating and validating novel technologies in hardware spiking neural networks

“…Supplementary Table S1 (Section 1) provides a detail-enriched version of the data illustrated in Figure 4 , which facilitates the in-depth comparison of the technical solutions of each platform to the readers seeking a deeper understanding of the landscape of the numerical tools for SNN simulations. The list of platform appears in order, from the earliest release to the most recent, and it encompasses the following: GENESIS ( Bower and Beeman, 2007 ), XPPAUT ( Bard, 1996 ), NEURON ( Hines et al, 2020 ), NCS ( Drewes, 2005 ; Hoang et al, 2013 ), EDLUT ( Ros et al, 2006 ), NEST ( Gewaltig and Diesmann, 2007 ), CARLSim ( Niedermeier et al, 2022 ), NeMo ( Fidjeland et al, 2009 ), CNS ( Poggio et al, 2010 ), GeNN ( Yavuz et al, 2016 ), N2D2 ( Bichler et al, 2017 ), Nengo ( Bekolay et al, 2014 ), Auryn ( Zenke and Gerstner, 2014 ), Brian 2 ( Stimberg et al, 2019 ), NEVESIM ( Pecevski et al, 2014 ), ANNarchy ( Vitay et al, 2015 ), MegaSim ( Stromatias et al, 2017 ), BindsNET ( Hazan et al, 2018 ), DynaSim ( Sherfey et al, 2018 ), SPIKE ( Ahmad et al, 2018 ), LSNN ( Bellec et al, 2018 ), cuSNN ( Paredes-Valles et al, 2020 ), Slayer ( Shrestha and Orchard, 2018 ), RockPool ( Muir et al, 2019 ), SpykeTorch ( Mozafari et al, 2019 ), PySNN ( Büller, 2020 ), s2net ( Zimmer et al, 2019 ), sinabs ( Lenz and Sheik, 2020 ), DECOLLE ( Kaiser et al, 2020 ), Spice ( Bautembach et al, 2020 ), Spiking Jelly ( Fang et al, 2020 ), Sapicore ( Moyal et al, 2021 ), Norse ( Pehle and Pedersen, 2021 ), Lava ( Richter et al, 2021 ), snnTorch ( Eshraghian et al, 2021 ), EvtSNN ( Mo and Tao, 2022 ), and Doryta ( Cruz-Camacho et al, 2022 ).…”

Enabling Efficient On-Edge Spiking Neural Network Acceleration with Highly Flexible FPGA Architectures

“…They may reduce the amount of processing since spike events are supposed to be sparse [18]. Clock-driven methods, however, are widely used because they can be more easily described and simulated [19], and many neuron models in this class can be constrained to follow these rules, which are meant for simulators that run through every time step.…”