russelljjarvis
commented
1 year ago I wonder if there is any problems if I just create this file for the repository: https://citation-file-format.github.io/
cff-version: 1.2.0 message: "If you use this software, please cite it as below." authors:
- family-names: Yao
given-names: Lu title: "Spiking Neural Networks" version: 0.1.0 doi: date-released:
Hi there SNN.jl code contributors (@AStupidBear @jpsamaroo @sosiristseng). I recently emailed you all about how to cite this repository, as I am using this code base as a foundation for other work, the idea of using this repository as a foundation raises the question: should I try to do science based on a fork, or PR back to this repository?
Rather than making a big PR, with all of my proposed changes, I decided to just make GH gists to share the code ideas I am talking about, this might be a more effective way to test consensus (or absence of consensus) about the proposed changes as a warm up before doing a PR.
Code Change Proposals:
The following gist is a proposed Potjan's and Diesmon connectome code that I am close to implementing with a very hacked version of
SNN.jl. Note I have discussed this wiring scheme in the preprint link: https://gist.github.com/russelljjarvis/107693cbf968421a2edf40a0b6f1ad63 A reference to Potjan's and Diesmon's model: https://pubmed.ncbi.nlm.nih.gov/23203991/A gist of the LIF cell population struct, where I propose changes to the struct to make the struct more GPU compliant. https://gist.github.com/russelljjarvis/3ac699d028f29f1e4f7d5052196930ba
A related gist function where, updates in main.jl are re-written to draw
V_{M}update and synaptic updates close together: https://gist.github.com/russelljjarvis/7a95b19693d6fdeb385f056d976811efMoving VM updates and weight updates to inside the same loop in the same update function would be similar to the Nvidia Cuda kernel approach inspired by Ben Arthur here: https://github.com/SpikingNetwork/TrainSpikingNet.jl/blob/master/src/gpu/loop.jl#L1-#L42
A gist file of code to do read the NMNIST data set into Julia via
tonic, making it possible to play event based camera data into SNN.jl spiking network models. Note 2/3rds of the way down the gist the language changes from Julia to Python @yeshwanthravitheja from my lab was largely responsible for this part.https://gist.github.com/russelljjarvis/ab7d76c4e5f061e038bd9487234abcf7 .
Propsed pre-print, that SNN code authors are welcome to add themselves to if they consent with this. https://typst.app/project/wuznt8iqEfRJWuX9zC9-eR (revised link, previous document broken)
Other Ideas.
SpikingNeuralNetworks.jltoSpikeTime.jl. I just thought SpikeTime sounds less like a literal description, potentially a bit more fun (ie "it's spike time!"), and SpikeTime still has a reference to temporal coding etc.If I can resolve all of objections to these ideas, I will create a PR that tries to merge these changes into the repositories.