ReparoSim

🎮 ReparoSim — Interactive Reparodynamics Simulator (TGRM vs Retrain)

Author: Cody R. Jenkins — Open Science Reparodynamics Initiative
Version: v1.0 (October 2025)
License: MIT

ReparoSim is a browser-based demo of Reparodynamics. It shows how the Targeted Gradient Repair Mechanism (TGRM) restores a degraded system under an energy budget, compared to a heavier retrain baseline.

▶️ Run the Simulator (GitHub Pages)

Live: https://bonemantgrm.github.io/ReparoSim/simulator/

If the link 404s, wait a minute for Pages to build or see “Enable GitHub Pages” below.

What you’ll see

• Accuracy over time (recovery)
• Energy usage (proxy)
• BPI (Bounded Power Index = performance gain / energy)
• Final Accuracy vs Energy (scatter)

Controls

• Steps (simulation length)
• Fault level (amount of degradation/noise)
• Top-K% (fraction of parameters TGRM repairs)
• k_repair (repair pulse intensity)

Enable GitHub Pages (once)

1. Repo → Settings → Pages
2. Build and deployment: “GitHub Actions”
3. Wait ~1 minute. Your simulator will be live at
   https://bonemantgrm.github.io/ReparoSim/simulator/

(This repo already includes a workflow in .github/workflows/pages.yml to auto-build Pages.)

Local preview (optional)

Just open simulator/index.html in a browser (no server needed).

Theory (very short)

• TGRM targets the most salient components (top-K%) and applies bounded repair pulses with diminishing returns.
• We track energy as a proxy (magnitude of updates/effort) and optimize BPI (stability gained per energy).

Contributing

PRs welcome! Add new fault modes, visualizations, or export buttons (CSV/PNG).

© 2025 Cody R. Jenkins — Open Science Reparodynamics Initiative. MIT License.