Russell Jarvis russelljjarvis

Hi @rafaqz

I reimplemented the Potjan's Diesmann synaptic wiring rules from Python (https://github.com/nest/nest-simulator/blob/master/pynest/examples/Potjans_2014/network_params.py) to Julia below.

The result actually works with the SNN package, we were talking about in Signal.

Its supposed to wire up to millions of cells, it comfortably wires 500,000 cells in less than 0.4seconds, but for 5,000,000 it takes an hour to create the matrix on a machine with 64GB of RAM.

Hi Raf

For a while I have been trying to massage this Python model into Julia, in part unsuccessfully.

https://portal.brain-map.org/explore/models/mv1-all-layers

The model is interesting for your grid simulator, as the neurons are arranged in pixel like grids (cells have both cartesian pixel indexs, and spatial coordinates).

Spike Time

Subtitle: Exploiting Modern Language Features for High Throughput Spiking Network Simulations at a Lower Tech Debt

Tags: Simulation of Spiking Neural Networks, Computational Neuroscience, Large Scale Modelling and Simulation

authors:

* Undecided order, or all equal order, I am flexible about this.
For example, in order that other other authors agree with,
name: Russell Jarvis affiliation: International Centre for Neuromorphic Systems, MARCS Institute, Western Sydney University

name: Yeshwanth Bethi affiliation: "International Centre for Neuromorphic Systems, MARCS Institute, Western Sydney University

	using SparseArrays
	#using ProgressMeter
	using UnicodePlots
	function potjans_params(scale_N_cells=1.0)
	"""
	This code draws heavily on the PyNN OSB Potjans implementation code found here:
	https://github.com/OpenSourceBrain/PotjansDiesmann2014/blob/master/PyNN/network_params.py#L139-L146
	"""
	conn_probs = [[0.1009, 0.1689, 0.0437, 0.0818, 0.0323, 0., 0.0076, 0. ],
	[0.1346, 0.1371, 0.0316, 0.0515, 0.0755, 0., 0.0042, 0. ],

	sing LinearAlgebra, Random, Statistics, StatsBase, ArgParse, SymmetricFormats
	using Plots
	using StatsBase
	using Random
	using ProgressMeter
	using StatsPlots
	using Plots
	#using UnicodePlots
	unicodeplots()

	include("src/genWeightsPotjans.jl")

	ccu = Dict("23E"=>20683,
	"4E"=>21915,
	"5E"=>4850,
	"6E"=>14395, "6I"=>2948, "23I"=>5834,"5I"=>1065,"4I"=>5479)
	ccu = Dict((k,ceil(Int64,v)) for (k,v) in pairs(ccu))

	Ncells = sum([i for i in values(ccu)])+1

	Hi @ali-mehrabi,

	To automatic docker building in github. Click on the tab "actions" in your repository.

	Then you can create an action similar to this:

	https://github.com/pabloabur/brian2-sims/blob/main/.github/workflows/docker-image.yml#L18-L20

	lines 18-20 build and run the dockerfile.

	"""
	I destroyed the generalness of the previous typing system which only worked for mutable kwarg structs

	This approach works but with unsophisticated typing.
	"""


	abstract type AbstractIFParameter end
	struct IFParameter# <: AbstractIFParameter
	τm::Float32# = 20ms

	using SparseArrays
	using StaticArrays
	using ProgressMeter
	using LinearAlgebra

	"""
	This file contains a function stack that creates a network with Potjans and Diesmann wiring likeness in Julia using SpikingNeuralNetworks.jl to simulate
	electrical neural network dynamics.
	This code draws heavily on the PyNN OSB Potjans implementation code found here:
	https://github.com/OpenSourceBrain/PotjansDiesmann2014/blob/master/PyNN/network_params.py#L139-L146


	using Revise
	using CUDA
	"""
	# * Changes to the cells struct as follows:

	# * Make the typing used dynamic and parametric, ie types can be either CuArray or regular Array, depending the input
	# arguments that are passed to the constructor.

	# TODO, I have good things about KernelAbstractions.jl it possible that kernel abstractions will remove my reliance on different method dispatch to choose CPU/GPU backend.