GPU SNN Simulator

This site is intended to give an overview and guide to the use of SPIKE, our Spiking Neural Network (SNN) Simulator.

The Overview page describes how the simulator is constructed and should give an intuition for the design principles behind Spike. Please see the Simple Example to get started. To create more complex models, see the Model Component sub-pages.

Spike is regularly used and tested on Linux and OS X (+ MacOS) machines with NVIDIA GPUs (Compute Capability >= 5.2). It is not tested and is not expected to work in Windows. It also requires the installation of CUDA (version >= 7.5) and CMAKE (>= v3.1).

See the most up to date version of Spike here:

https://github.com/nasiryahm/Spike/

Building and Running Spike

Spike comes with a few example scripts which are located in the Examples folder of the repository.

To build these Examples:

Download or clone the repository
In a terminal, navigate to the build directory
Run the bash script

On a linux machine, this boils down to these commands:

git clone https://github.com/nasiryahm/Spike/

cd Spike/Build/

./build.sh

Running the Examples

To run the VogelsAbbott example:

cd Spike/Build/Examples

./VogelsAbbottNet

This example will output two text files with the spike ids and spike times for this simulation.

Speedy Spike

Spike is currently one of the fastest SNN simulators available.

The graphs right show the run-time of Spike per second of simulation against other simulators. Note that the y-axes are log scale. Smaller is better.

The upper graph shows the Vogels-Abbott Benchmark: a comparison using a 4000 Leaky-Integrate and Fire Neuron Network with 2% connectivity.

The lower graph shows the Brunel Benchmark: a comparison using a 10,000 Leaky-Integrate and Fire Neuron Network with 10% connectivity (10^7 synapses). Red bars show the same network when plasticity is active. Black bars are the network with no plasticity.

For more information and comparisons of Spike to other SNN simulators, see:

https://github.com/nasiryahm/SNNSimulatorComparison

Vogels-Abbott Benchmark

Brunel Benchmark