Light

yjwu17 / stbp-for-training-spikingnn Goto Github PK

View Code? Open in Web Editor NEW

203.0 6.0 48.0 23.49 MB

Spatio-temporal BP for SNNs

Python 100.00%

stbp-for-training-spikingnn's Introduction

Spatio-temporal BP for spiking neural networks.

Matlab version of convolutional SNN on MNSIT[1].

Please find another branch for Pytorch version on CIFAR10[2].

For neurmorphic dataset(N-MNIST and DVS-Gesture), please refer to examples of our another projects[3]:

https://github.com/hewh16/SNNs-RNNs

Requirement

Python 3.6
MNIST dataset
CIFAR10 dataset
N_MSNIT dataset

Results

After 100 epochs, it can obtain ~ 99.4% acc on MNIST.

Reference

Wu, Yujie, Lei Deng, Guoqi Li, Jun Zhu, and Luping Shi. "Direct Training for Spiking Neural Networks: Faster, Larger, Better." arXiv preprint arXiv:1809.05793 (2018).
Wu, Yujie, Lei Deng, Guoqi Li, Jun Zhu, and Luping Shi. "Spatio-temporal backpropagation for training high-performance spiking neural networks." Frontiers in neuroscience 12 (2018).
He W, Wu Y J, Deng L, et al. Comparing SNNs and RNNs on neuromorphic vision datasets: Similarities and differences[J]. Neural Networks, 2020, 132: 108-120.

stbp-for-training-spikingnn's People

Contributors

Stargazers

Watchers

Forkers

arthas1121 treestreamymw fuqianggu relevation-143 mashijie1028 hmiku-suki seaeastxu fdusnngroup wangyang24 haoyz rachmadvwp 1895-art xdusponge toanhvu jpjp927 shanka19 xiaoanh lemonade117 gavinatthu kanaderu githubmalajava jxy5969 thegoist qianpengli577 ericgreatgeek wisamreid iridescentmay rafaelblevin821 wzb1005 201528014227051 hanxiao7533 lucazanatta zhaxidele xgkiri zhyan567 xiaocuilin quanztx zfd-1 ccchaosss sirius-johnny alvaro-zs unicornshell nanjunye yqlove1 caohlin bhkim003 m0myhudz

stbp-for-training-spikingnn's Issues

how to implement the encoding scheme?

hello, I want ask a question about "Direct Training for Spiking Neural Networks: Faster, Larger, Better“.
How to implement the encoding scheme. Can you give some details. Thanks!

cfg_kernelsize

Hi,

I am referring to your code. My question is how you are taking cfg_kernel_size?

I didn't get the forward function, how you kept cfg_kernel_size? Below is the snippet of your forward function.

Can you please tell me how did you do that?

torch.zeros(batch_size, cfg_cnn[0][1], cfg_kernel[0], cfg_kernel[0], device=device))

Thanks

Derivative Approximation of the Non-Differentiable Spike Activity

Should the temp variable be divided by 2 * lens (i.e temp = (abs(input - thresh) < lens) / (2 * lens) ) ? You choose lens = 0.5 in your code so there is no problem, but if someone wants to tune this parameter the derivative approximation is not correct (based on the paper "Spatio-Temporal Backpropagation for Training High-Performance Spiking Neural Networks").

cifar10 code，loss remains unchanged

about the pytorch version for implementing cifar10 tasks ，sorry i don't know where the bug is，loss remains unchanged，low accuracy

Reference 1

Can you provide the code of NeuNorm in Reference 1?

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.