Skip to content

About

Collection of examples from the JuliaSNN organization

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

SNNExamples

Runnable examples of the JuliaSNN packages: single neurons, networks, plasticity, rate models, perturbations, extensions of the model library, and re-implementations of published models.

JuliaSNN is a set of Julia packages to simulate networks of spiking neurons (point and dendritic neuron models, conductance and receptor synapses, delays, long- and short-term plasticity, Poisson and spike-time inputs) and to record, analyse and plot the results:

Package Role
SNNModels.jl models and simulation engine
SpikingNeuralNetworks.jl loads and re-exports the packages below (using SpikingNeuralNetworks, alias SNN)
SNNPlots.jl plots of recordings (Makie)
SNNUtils.jl stimulation protocols and analysis

Documentation: https://juliasnn.github.io/SpikingNeuralNetworks.jl. The examples run with SNNModels 2.1.0, SpikingNeuralNetworks 2.0.1, SNNUtils 1.0.0 and SNNPlots 1.0.0 (see Project.toml).

This repository is not a Julia package and is not registered: Project.toml and Manifest.toml only pin the environment the examples run in. Clone it and instantiate the project as shown below.


Firing patterns of the AdEx neuron
tutorials/neurons/AdEx_neuron.jl

E/I network of AdEx and IF neurons: raster, traces, rates
tutorials/networks/AdEx_net.jl

STDP kernels of the long-term plasticity rules
tutorials/STDP_kernel.jl

Short-term plasticity: efficacy, resources and utilisation
tutorials/plasticity/STP_demo.jl

Perturbing a running E/I network
tutorials/perturbations/EI_network_perturbation.jl

Rate network trained to follow a target (PINning)
tutorials/rate_models/pinning_net.jl

Tripod neuron: soma, dendrites and receptor conductances
tutorials/neurons/Tripod_neuron.jl

COBA benchmark network
tutorials/networks/COBA.jl

Run an example

git clone https://github.com/JuliaSNN/SNNExamples.git
cd SNNExamples
julia --project=. -e 'using Pkg; Pkg.instantiate()'       # first time only, a few minutes

julia --project=. tutorials/neurons/AdEx_neuron.jl         # run a script
julia --project=. scripts/run_example.jl tutorials/neurons/AdEx_neuron.jl   # headless: saves the figures to figures/

Most scripts end with a Makie figure. In the Julia REPL or in VS Code include("tutorials/neurons/AdEx_neuron.jl") shows it; scripts/run_example.jl runs the script with the CairoMakie PNG backend and writes every figure it creates to figures/ (use --figdir DIR and --max-figs N to change this). Each script starts with a comment that says what it shows and how long it takes (the time includes Julia start-up and compilation, on a 24-core workstation).

The tutorials follow the pages of the documentation (populations, plasticity, simulations).

Layout

Folder Content
tutorials/ neurons/, networks/, plasticity/, rate_models/, perturbations/, extensions/, analysis/, usage/ (saving and loading models, the graph of a model, a SLURM job), and a few scripts at the top level (spike-time stimulus, STDP)
experiments/ re-implementations of published models: Mongillo 2008, Litwin-Kumar & Doiron 2014, Tsodyks 1997, Potjans & Diesmann 2014, Zerlaut 2019, Festa 2024, competing assemblies, and others, plus parameters/ (parameter sets used by other projects); some are not runnable, see STATUS.md
scripts/ run_example.jl (headless runner), run_all.jl (runs everything with timeouts), make_status.jl, examples.toml (the list of examples)
.github/workflows/ CI: smoke run of a few fast examples
figures/ PNG figures written by the runner
STATUS.md result of the last full run: pass, fixed, not run, with the reason

Examples

Status: pass = ran as it came from SpikingNeuralNetworks.jl, fixed = ported to the current API and now runs, not run = not executed (cluster data, removed models, unfinished code; reason in STATUS.md). "Ran" means the script finished without error; STATUS.md also lists observations about results that look wrong (for example the Izhikevich tutorials).

Tutorials: neurons

Example What it shows Status
tutorials/neurons/IF_neuron.jl IF neuron under excitatory and inhibitory Poisson input pass
tutorials/neurons/IF_noise_input.jl IF neuron under a noisy input current pass
tutorials/neurons/AdEx_neuron.jl AdEx firing patterns under a current step (Naud et al. 2008) pass
tutorials/neurons/AdEx_threshold.jl AdEx neuron with an adaptive spike threshold pass
tutorials/neurons/AdEx_balanced_stimuli.jl AdEx neuron in the high-conductance state, balanced Poisson input pass
tutorials/neurons/AdEx_multireceptors.jl AdEx neuron with AMPA/NMDA/GABA-type receptors of different kinetics pass
tutorials/neurons/Izikievich_neuron.jl Izhikevich firing patterns pass
tutorials/neurons/HH_neuron.jl Hodgkin-Huxley neuron, f-I curve pass
tutorials/neurons/MorrisLecar_neuron.jl Morris-Lecar neuron, noisy current and phase plane pass
tutorials/neurons/BallAndStick_neuron.jl Ball-and-stick neuron, receptor conductances pass
tutorials/neurons/Tripod_neuron.jl Tripod neuron (soma and two dendrites), AMPA/NMDA/GABA receptors pass
tutorials/neurons/Tripod_multireceptors.jl Tripod neuron with three GABA receptors pass
tutorials/neurons/tripod2.jl Tripod neuron under dendritic Poisson input pass
tutorials/neurons/HetRec_layer.jl HetRec layer with heterogeneous dendritic time constants pass
tutorials/neurons/PoissonPop.jl Poisson population: raster, rates, ISI CV2 pass
tutorials/timed_stim.jl Spike-time stimulus pass

Tutorials: networks

Example What it shows Status
tutorials/networks/IF_net.jl Balanced IF network (3200 E, 800 I) pass
tutorials/networks/HH_net.jl Hodgkin-Huxley network (no figure) pass
tutorials/networks/Izikievich_net.jl Izhikevich network pass
tutorials/networks/AdEx_net.jl AdEx/IF network with heterogeneous parameters and delays pass
tutorials/networks/AdExVarStimuli.jl Network driven by a time-varying Poisson rate pass
tutorials/networks/COBA.jl COBA benchmark network (Brette et al. 2007) pass
tutorials/networks/CUBA.jl CUBA benchmark network (Brette et al. 2007) pass
tutorials/networks/CUBA_brian.py CUBA benchmark in Brian2, for the speed comparison not run
tutorials/networks/afferent_response.jl Network rate against afferent rate (Zerlaut et al. 2019 parameters) pass
tutorials/networks/weak_strong_afferents.jl Network synchrony against afferent rate pass
tutorials/networks/ParadoxicalEffect.jl Paradoxical response of an inhibition-stabilised network (Tsodyks et al. 1997) pass
tutorials/networks/passive_vs_active_dendrites.jl Networks of passive and NMDA-active dendrites pass
tutorials/networks/out_degree.jl Out-degree distributions of the connectivity rules pass
tutorials/networks/chain.jl Feed-forward chain of three IF neurons fixed

Tutorials: plasticity

Example What it shows Status
tutorials/plasticity/STP_demo.jl Short-term plasticity of one synapse: ρ, u, x pass
tutorials/networks/AdEx_STP.jl Network with short-term depression pass
tutorials/networks/STP_het.jl Heterogeneous short-term plasticity from experimental fits pass
tutorials/stdp_demo.jl Pair STDP: weight change against spike-time difference pass
tutorials/STDP_kernel.jl Kernels of all long-term plasticity rules pass
tutorials/plasticity/AdEx_plasticity.jl vSTDP network with synaptic normalisation pass
tutorials/networks/synaptic_turnover.jl Random synaptic turnover pass
tutorials/plasticity/AdEx_turnover.jl Activity-dependent turnover in space pass
tutorials/plasticity/aggregate_scaling.jl Homeostatic aggregate scaling towards a target rate fixed

Tutorials: rate models

Example What it shows Status
tutorials/rate_models/rate_net.jl Rate network pass
tutorials/rate_models/Oja_rule.jl Oja rule in a rate network pass
tutorials/rate_models/pinning_net.jl PINning: recurrent network trained on a target pass
tutorials/rate_models/ppc.jl Rate network with a trained readout pass

Tutorials: perturbations, extensions, analysis

Example What it shows Status
tutorials/perturbations/EI_network_perturbation.jl Perturbing a running E/I network pass
tutorials/extensions/neuron_model.jl Defining a new neuron model pass
tutorials/extensions/stimulus_model.jl Defining a new stimulus model pass
tutorials/extensions/doubleExpCurrent.jl Defining a new synapse model pass
tutorials/analysis/recordings.jl Every way to read a recording pass
tutorials/analysis/simulation_speed.jl Simulation speed of AdEx, Tripod and BallAndStick (BenchmarkTools) fixed

Tutorials: usage

Example What it shows Status
tutorials/usage/load_model.jl Save a network, load it, continue it, rerun it without one stimulus pass
tutorials/usage/get_graph.jl Graph of a model (populations, synapses, stimuli) fixed
tutorials/usage/load_cluster.jl SLURM job: one network per worker process, saved to disk fixed

Experiments: papers

Example What it shows Status
experiments/Mongillo2008_WorkingMemory/STP_onecell.jl Mongillo et al. 2008: Tsodyks-Markram synapse, x and u after a spike train fixed
experiments/Mongillo2008_WorkingMemory/STP_WM_Mongillo2008.jl Mongillo et al. 2008: working memory by synaptic facilitation, 10,000 neurons fixed
experiments/OtherPapers/Tsodyks1997_ParadoxicalEffect.jl Tsodyks et al. 1997: paradoxical effect, 2000 AdEx + 500 IF neurons fixed
experiments/OtherPapers/LitwinKumar2014_vSTDPNetwork.jl Litwin-Kumar & Doiron 2014: balanced AdEx network with their parameters fixed
experiments/Festa2024_StructuredInhibition/structured_inhibition_symmetric_STDP_network.jl Festa et al. 2024: structured inhibition with symmetric STDP on I -> E (550 s) fixed
experiments/Festa2024_StructuredInhibition/structured_inhibition_asymmetric_STDP_network.jl Festa et al. 2024: structured inhibition with anti-symmetric STDP on I -> E (650 s) fixed
experiments/LitwinKumar2014_AssociativeMemories/simulation.jl Litwin-Kumar & Doiron 2014: network with vSTDP, iSTDP and normalisation, 3 s fixed
experiments/PotjansDiesmann2014_CorticalCircuit/multilayer_potjans-diesmann2015.jl Potjans & Diesmann 2014: 8-population cortical microcircuit at 1 % scale fixed
experiments/Zerlaut2019_SpectrumAsynchronous/simulation_example.jl Zerlaut et al. 2019: raster and rates of the 10,000-neuron network at 4 and 10 Hz fixed
experiments/Zerlaut2019_SpectrumAsynchronous/simulation_standalone.jl Zerlaut et al. 2019: transfer function, 4000 E + 1000 I, 20 rates x 5 runs fixed
experiments/Zerlaut2019_SpectrumAsynchronous/simulation.jl Zerlaut et al. 2019: transfer function, 8000 E + 2000 I, 20 rates x 5 runs fixed
experiments/tripod_soma.jl Tripod neuron: response to one excitatory and one inhibitory somatic spike fixed

Experiments: competing assemblies (BistablePopulations)

Example What it shows Status
experiments/BistablePopulations/two_competing_populations.jl Two E/I networks competing through lateral excitation of inhibition fixed
experiments/BistablePopulations/bistable_network_with_stimulus.jl Two assemblies and an alternating decaying stimulus fixed
experiments/BistablePopulations/multistable_network_with_stimulus.jl Four assemblies with inhibitory plasticity, two cued in turn fixed
experiments/BistablePopulations/persistent_adapting_activity.jl Persistent activity with four inhibitory plasticity rules (40 s each) fixed

Experiments: not run

Example What it shows Status
experiments/Lagzi2022_AssemblyFormation/simulation_stdp_baseline.jl Lagzi et al. 2022: STDP training of the assembly network over a grid of stimulus rate x timescale x SST count not run
experiments/Lagzi2022_AssemblyFormation/simulation_stdp_run.jl Lagzi et al. 2022: single STDP training run not run
experiments/Lagzi2022_AssemblyFormation/simulation_stdp_test.jl Lagzi et al. 2022: 5 s test of the network with NMDA and AMPA receptors not run
experiments/Lagzi2022_AssemblyFormation/simulation_stdp_load.jl Lagzi et al. 2022: load trained networks and analyse the weights not run
experiments/Lagzi2022_AssemblyFormation/simulation_stdp_load_mixed.jl Lagzi et al. 2022: load trained networks with mixed inhibition not run
experiments/Lagzi2022_AssemblyFormation/simulation_stdp_baseline_analysis.jl Lagzi et al. 2022: analysis of the baseline grid not run
experiments/Lagzi2022_AssemblyFormation/simulation_auditory_stim_response.jl Lagzi et al. 2022: response of trained networks to recorded sounds not run
experiments/Lagzi2022_AssemblyFormation/simulation_auditory_stim_response_analysis.jl Lagzi et al. 2022: analysis of the sound responses not run
experiments/Lagzi2022_AssemblyFormation/auditory_spikes.jl Lagzi et al. 2022: Poisson spikes sampled from auditory cortex rates not run
experiments/Lagzi2022_AssemblyFormation/autocor.jl Lagzi et al. 2022: autocorrelation of the Ornstein-Uhlenbeck rate signal not run
experiments/Lagzi2022_AssemblyFormation/autocor_STDP.jl Lagzi et al. 2022: autocorrelation of the signal after STDP not run
experiments/Vlachos2025_Sequences/simulation.jl Vlachos et al. 2025: sequences in a network of multi-timescale AdEx neurons not run
experiments/Sarazin2022_NeuralSequences/if_CANAHP_net.jl Sarazin et al. 2022: IF network with CAN and AHP currents not run
experiments/Seeholzer2019_RingAttractor/simulation.jl Seeholzer et al. 2019: ring attractor network not run
experiments/Seeholzer2019_RingAttractor/optimization.jl Seeholzer et al. 2019: parameter optimisation on a cluster not run
experiments/PotjansDiesamann2014_network/potjans_nmnist.jl Potjans-Diesmann network driven by N-MNIST event data not run
experiments/PotjansDiesamann2014_network/batch_nmnist_motions.py N-MNIST batch conversion not run

Experiments: parameters

Example What it shows Status
experiments/parameters/dendritic_WM.jl Parameters/network function for a working-memory model of ball-and-stick neurons not run
experiments/parameters/dendritic_network_phoncat.jl Parameters of the dendritic networks (Tripod/ball-and-stick, PV, SST) of the phoneme-categorisation project not run

Check everything

julia --project=. scripts/run_all.jl            # every example in its own process, at most 6 in parallel, 300 s timeout
julia --project=. scripts/run_all.jl tutorials/neurons --jobs 2   # a folder or single scripts
julia --project=. scripts/make_status.jl        # rewrites STATUS.md and the table above

Logs are written to _logs/ (ignored by git).

Adding an example

  1. Put the script in the matching folder of tutorials/ or experiments/. Start it with a comment that says what it shows and how long it takes, for example # Runtime: about 40 s. (look at any script in tutorials/neurons/).
  2. Add an [[example]] entry to scripts/examples.toml (path, topic, desc, status, reason, changed, origin, note).
  3. Check it: julia --project=. scripts/run_example.jl <path>; the last output line is RESULT pass|fail ....
  4. Run julia --project=. scripts/make_status.jl to rewrite STATUS.md and the table above.

Notes

  • SNNModels 2.1.0 changed the Tripod and BallAndStick defaults (soma clamp, GABA reversal potentials, dendrite diameter); results of the Tripod examples differ from earlier versions. See the SNNModels changelog.
  • Upgrading code from version 1 of the packages: What changes from v1 to v2.
  • License: MIT, see LICENSE.

About

Collection of examples from the JuliaSNN organization

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages