MAV-MCC Generale (vNext): Boundary Bridge + Grav-AB No-Fit + Class-Scaling (FREEZE)
MAV-MCC General Theory — v4.0 (FREEZE): Boundary Bridge (L0+EM + boundary action) with strict NO-FIT governance and Class-Scaling (NO-REFIT) per material class.Includes external validation tests: Chang 2012 (Au–Au), Laurent 2012 (Au–Au vs Au–Si, 4.2 K), Liu 2021 (graphene thermal case), digitized into reproducible CSV tables.Continuity: new version of 10.5281/zenodo.18289899; supplement to 10.5281/zenodo.18070901.::contentReference[oaicite:0]{index=0} Release v4.0 (FREEZE): integrates external tests (Chang 2012, Laurent 2012, Liu 2021) and formalizes Class-Scaling (NO-REFIT) per material class.MAV-MCC General Theory — v4.0 release. ## Abstract / ScopeThis release formalizes MAV-MCC Generale (vNext) as a restart-grade, audit-ready package.It consolidates the “Boundary Bridge (L0+EM + boundary action)” scaffold, enforces strict NO-FIT governance (incl. null-test logic), and introduces Class-Scaling (NO-REFIT) as a predictive layer by material class. ## What’s inside (source-of-truth)- Canonical master table (CANON): datasets, parameters (A,B,S), PASS/FAIL metrics, and links to “truth” CSVs- External validation set #1–#3 (digitized into reproducible CSV tables): - Chang 2012: Au–Au AFM (force-gradient) - Laurent 2012: Au–Au vs Au–Si (cryogenic) - Liu 2021: graphene thermal case- Deterministic rules: - PASS-95 criterion - normalized residual r95 - s1 sign rule (NO-FIT) - Class-Scaling rules per class (C1/C2/C3)- Verification scripts for quick reproducibility checks ## Key claims (operational, test-based)- Boundary-law robustness across multiple materials (PASS-95 in frozen truth datasets)- Exit from post-hoc tuning via strict governance and independent null-test logic- Class-Scaling: one-time class calibration, then NO-REFIT predictions on external datasets (audit trail included) ## How to use1) Download the ZIP2) Start from the SPEC3) Treat CANON as source-of-truth4) Run verification scriptsAll external tests are included as CSV tables under `data/external_tests/`. Continuity- Is new version of: 10.5281/zenodo.18289899- Is supplement to: 10.5281/zenodo.18070901 How to cite:De Franco, Gianni. MAV-MCC Generale (vNext): Boundary Bridge + Grav-AB No-Fit + Class-Scaling (FREEZE) (v4.0). Zenodo. https://doi.org/10.5281/zenodo.18367044
Authors
- Gianni De Franco
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-01-25
- DOI
- https://doi.org/10.5281/zenodo.18367044
- Primary Topic
- Machine Learning in Materials Science
- Type
- article
- Field-Weighted Citation Impact
- 0.00