SNN-lab

Experiments of SNN (Spiking Neural Network), focusing on several metrics, e.g. sparsity, accuracy. Most workflows show 80 to 90 percent of sparsity (10% ~ 20% neurons are activated on average)

You can directly run any scripts and do tweaks, once the env is setup.

This repo is only tested on Ubuntu 24.04, CPU + GPU, for simulation only. For real efficiency experiment like throughput and power-efficiency, SNNs need to run on neuromorphic hardware.

🚀 welcome to open issues and discussions in "Issue"

Quick start

Setup env, and run

python snn-torch/mnist_linear_sparsity.py # or any other path to the script

Descriptions

• snn-torch: experiments using snnTorch library, some code are adapted from snnTorch tutorial

Env setup

sudo apt-get install python3-tk  # for matplotlib backend
pip install snntorch