A CLI and library for computing Φ_spectral (spectral integrated information) over time-series data such as EEG recordings and simulated neural networks.
Based on the work of Mark M. Bailey:
- Source: Spectral-Integrated-Information
- Papers: "When Wholes Resist Decomposition" (Bailey & Schneider, 2025); "Superpsychism" (Schneider & Bailey, forthcoming JCS)
# Bootstrap the project (creates venv, installs dependencies)
bash phi_spectral_bootstrap.sh
# Activate the environment
source .venv/bin/activate
# Run the full pipeline
bash scripts/run_pipeline.sh --simulate-all
# Compute Φ_spectral on your own data
phispectral compute --input data/raw/eeg.csv --window 10 --step 1 --bins 3 --outdir runs/my_run
# Parameter sweep
phispectral sweep --input data/raw/eeg.csv --windows 5,10,20,50 --overlay --errorbars --outdir runs/sweepPlace input files in data/raw/. Supported formats:
.csv— wide format (rows = time, columns = nodes).json.npy— shape(n_nodes, T)
| Parameter | Value |
|---|---|
| n (nodes) | 50 |
| T (timesteps) | 150 |
| window | 10 |
| step | 1 |
| bins | 3 |
CTLN simulation: θ=1.0, ε=0.25, δ=0.5, dt=0.1
| File | Description |
|---|---|
phi_timeseries.csv |
Time-centered Φ_spectral values (raw and normalized) |
phi_timeseries.png |
Φ_spectral over time plot |
traces.png |
Input signal traces |
mi_final.png |
Mutual information matrix heatmap (final window) |
mi_final.npy |
MI matrix as NumPy array (with --save-mi-final) |
| File | Description |
|---|---|
window_N/phi_timeseries.csv |
Per-window Φ time series |
window_N/phi_timeseries.png |
Per-window Φ plot |
phi_overlay.png |
All window sizes overlaid (with --overlay) |
phi_vs_window.png |
Mean Φ vs window size (with optional error bars) |
phi_vs_window.csv |
Mean and std Φ per window size |
Generates synthetic time-series data (.csv or .npy) for models: random, transitional, synchronized, ctln. CTLN can also save the network graph as .gml with --save-graph.
MIT — see LICENSE. Core library by Mark M. Bailey.