FEEL-SNN

Code for "FEEL-SNN: Robust Spiking NeuralNetworks with FrequencyEncoding and Evolutionary LeakFactor (NeurIPS 2024)"

Prerequisites

The Following Setup is tested and it is working:

• Python>=3.5

• Pytorch>=1.9.0

• Cuda>=10.2

Data preparation

• CIFAR10: def build_cifar(use_cifar10=True) in data_loaders.py

• CIFAR100: def build_cifar(use_cifar10=False) in data_loaders.py

• Tiny-ImageNet:

  (1) Download Tiny-ImageNet dataset

  (2)def build_tiny_imagenet() in data_loaders.py

Description

• Use a triangle-like surrogate gradient ZIF in layers.py for step function forward and backward.

• Use FE method def ft(x,freq_filter) in layers.py.

• Use EL employed spiking neuron LIFSpikeTau in layers.py.

FEEL Training & Testing

• Script for Vanilla+FEEL

  train: run bash script/vanilla_feel.sh;

  test: run bash script/vanilla_feel_test.sh.

• Script for AT+FEEL

  train: run bash script/at_feel.sh;

  test: run bash script/at_feel_test.sh.

• Script for RAT+FEEL

  train: run bash script/rat_feel.sh;

  test: run bash script/rat_feel_test.sh.

Citation

@article{xu2024feel,
  title={FEEL-SNN: Robust Spiking Neural Networks with Frequency Encoding and  Evolutionary Leak Factor},
  author={Xu, Mengting and Ma, De and Tang, Huajin and Zheng, Qian and Pan, Gang},
  journal={Advances in Neural Information Processing Systems},
  year={2024}
}

Repository Contributor: Mengting Xu, Qian Zheng