LaNMM_predictive_coding_paper

Code associated with CFC plots in Ruffini et al. 2025: https://www.biorxiv.org/content/10.1101/2025.03.19.644090v3

LaNMM model reference: https://www.sciencedirect.com/science/article/pii/S105381192300085X?via%3Dihub

Repository layout

• julia/: current Julia implementation (engine, analysis, plots, sweep, run scripts).
• python/: legacy/original Python implementation (lanmmv11 stack).
• notebooks/: exploratory notebooks and figure-building workflows.
• julia_lanmm_sweep_*/: generated sweep artifacts currently tracked in repo history.

Quick start (2 minutes)

cd /path/to/LaNMM_predictive_coding_paper
julia --version
julia --threads auto julia/run_job_P1.jl

Install Julia (macOS / Windows / Linux)

Recommended on all platforms: juliaup (official installer + version manager).

macOS / Linux

curl -fsSL https://install.julialang.org | sh
julia --version

Windows (PowerShell)

winget install --id JuliaLang.Juliaup -e
julia --version

If julia is not found, restart your terminal/session.

Run the Julia implementation

From the repository root:

julia --threads auto julia/run_job_P1.jl
julia --threads auto julia/run_job_P2.jl
julia --threads auto julia/run_job_PV.jl

• julia/run_job_P1.jl: multiscale driving on P1, constant on P2/PV.
• julia/run_job_P2.jl: multiscale driving on P2, constant on P1/PV.
• julia/run_job_PV.jl: multiscale driving on PV, constant on P1/P2.

Julia architecture

• julia/LaNMM_Types.jl: typed config/result contracts.
• julia/LaNMM_parameters.jl: intrinsic constants and default driving model.
• julia/LaNMM_Engine.jl: ODE and single-simulation runtime.
• julia/LaNMM_Analyzer.jl: analysis metrics (SEC/EEC, power, PEIX, peaks).
• julia/LaNMM_Plots.jl: plotting only.
• julia/LaNMM_Sweep.jl: sweep orchestration, progress, saving.

Full Julia-specific walkthrough:

• Readme_Julia_implemenation.md

Outputs per run

Each run writes a timestamped folder containing:

• parameters.txt (full reproducibility snapshot)
• diagnostic plots (inputs.png, v.png, u_external.png, psd.png)
• sweep plots (couplings.png, power.png, peix.png, frequency_heatmaps.png)
• serialized data (sweep_results.jls, analysis_results.jls)

Notes

• First run is slower due to Julia package precompilation.
• For faster tests, reduce sweep ranges and tmax in julia/run_job_*.jl.
• Sweep output folders are currently versioned; very large .jls files can hit GitHub limits, so prefer archiving externally or using Git LFS for big runs.