kclip / bayesian-snn

Bayesian spiking neural networks (SNNs)

Code for training SNNs via Bayesian learning using Intel's Lava framework (https://lava-nc.org/). This code has been used for the following work:

N. Skatchkovsky, H. Jang, and O. Simeone, Bayesian Continual Learning via Spiking Neural Networks (https://arxiv.org/pdf/2208.13723.pdf)

Running examples

Scripts for the paper are all in the scripts folder. Example syntax to run the scripts are given in launch_scripts.sh.

Running scripts on the MNIST-DVS and DVSGestures dataset requires to use our neurodata data preprocessing and loading package available at https://github.com/kclip/neurodata.

Dependencies

lava-nc v.0.3.0

lava-dl v.0.2.0

numpy v.1.22.3

pytables v.3.6.1

scikit-learn v.1.1.0

torch v.1.11.0

tqdm v.4.64.0

Author: Nicolas Skatchkovsky