Canonical alignment notice (v10)

MPFST's canonical reference is the "MPFST Publication Manuscript (v10)". Earlier site material that framed MPFST around a mu–gamma–H coherence meter is preserved for transparency but is now archived as legacy and under reevaluation.

Read the canonical manuscriptReplication protocols (canonical)Legacy archive

Canonical replication

Replication protocols for experimentalists (canonical, v10)

Recommended workflows for testing MPFST using the canonical gate definition (meltdown threshold + meltdownFrac). Older protocols that compute mu, gamma, H and mℓ are preserved as legacy proxies and are not the canonical definition of coherence in MPFST.

Read canonical manuscript (v10) Legacy archive (mu–gamma–H)

Protocol A: Simulation-true gate state (recommended workflow)

Goal: derive the canonical gate state directly from simulated fields. Report all parameter files and solver settings; use the same gate definition across runs.

  1. Define the combined synergy amplitude (canonical):

    M(x,t) = u4(x,t) + u5(x,t) + u6(x,t) + u7(x,t) + u8(x,t) + d(x,t)

  2. Compute partial and full gate fractions (defaults, report what you use):

    partial_meltdownFrac(t) = fraction of domain where M(x,t) > 0.5*Mth

    full_meltdownFrac(t) = fraction of domain where M(x,t) > 0.8*Mth

  3. Classify system state using these thresholds (or documented alternatives):
    • No transition: partial = 0 and full = 0
    • Partial transition: partial > 0 and full = 0
    • Full transition: full > 0
  4. Minimum reporting: gating thresholds used, spatial/temporal resolution, solver settings, and any boundary/initial condition choices. Optional upgrades: sensitivity analysis across nearby thresholds; cross-check with structured nulls.

Protocol B: Experimental gate proxy (for real-world datasets)

Because u4..u8 and d are latent model fields, experimental data should use an explicit proxy variable S(t) designed to track "synergy amplitude" in a modality-appropriate way. Requirements vary by modality (EEG/MEG/LFP/optics/mechanics); deviations are reportable parameters, not failures.

  1. Define S(t) before analysis (pre-registration).
    • EEG: weighted sum of band-limited amplitude envelopes plus a phase-alignment term.
    • Plasma: edge/pedestal instability amplitude proxy plus fast transient detector.
    • GW strain: post-merger residual or echo-energy proxy in defined windows.
  2. Thresholding (tiered guidance):
    • Minimum: declare the threshold rule in advance (baseline window or physically motivated criterion).
    • Recommended: report sensitivity across a small, preregistered range.
    • Optional (high-precision): external calibration or disciplined clocking to validate timing/latency.
  3. Define proxy_meltdownFrac on a rolling window: proxy_meltdownFrac = fraction of samples (or TF bins) in window where S exceeds threshold. Report window length and hop size.
  4. Run structured nulls: phase-scrambled controls, time-shift controls, and surrogate data matched on power spectrum where appropriate.
  5. Minimum viable reporting: sampling rate and latency characteristics, preprocessing (filtering/referencing/artifact handling), threshold rule, and null controls used. Optional: hardware coherence validation and independent reruns across sites.

Replication standard

MPFST is intended to be evaluated by independent teams using transparent pipelines, structured nulls, and explicit preregistration where possible. Code outputs are test generators, not empirical proof. HPC tooling assists exploration of the PDE system but is not a magic predictor.

Modality-specific defaults can be adapted; changes should be logged and treated as parameters for comparison, not as protocol breaks.

Legacy protocol (archived)

The prior mu–gamma–H experimental protocol is archived for transparency and historical reproducibility. It is under reevaluation relative to the canonical meltdownFrac gate and should be treated as an observational proxy, not the definition of coherence.

View the legacy mu–gamma–H protocol