This repository contains Jupyter notebooks describing a workflow that could be used to re-create the figures in Arthur, Kim, Chen, Preibisch, and Darshan (2023). The actual figures were made using the Linux command line interface to TrainSpikingNet.jl to facilitate batching parallel jobs out to an on-premise cluster, instead of the Julia REPL as demonstrated here. Gadfly.jl was also used for the manuscript's figures, whereas here we use Makie.jl.
If you want to follow along, you'll need to install and run Jupyter:
julia> ]add IJulia
julia> using IJulia
julia> notebook()
The default kernel uses just one thread. To create a kernel that fully utilizes your CPU cores:
installkernel("Julia auto threads", env=Dict("JULIA_NUM_THREADS" => "auto"))
To use Jupyter remotely, on a headless machine for example:
julia> using IJulia
julia> IJulia.find_jupyter_subcommand("")[1] # copy the output for later
Then, on the Unix command line:
ssh <your-machine> -L 8888:localhost:8888
<path-to-jupyter-from-above> notebook --no-browser --port=8888
See JuliaLang/IJulia.jl#586 for more details.