graph-robustness-score 0.1.1

Simple scalar robustness score for networks: spectral gap − singular-value variance + Ollivier–Ricci curvature

Graph Robustness Score C(G)

One clean number that tells you how robust any network is.
Spectral gap − singular-value variance + real Ollivier–Ricci curvature

Live Demo • Paper (Feb 19, 2026) • Full package + experiments

I offer paid network robustness audits (supply-chain, transaction graphs, biotech interaction networks, cyber attack surfaces).
First 3 clients this month get 30% off.

---

Why this score?

Most graph metrics only look at one thing (connectivity, clustering, or degree distribution).
C(G) combines three complementary views in a single interpretable number:

• Spectral gap → how well the whole network is connected (big gap = harder to disconnect)
• Singular-value variance → penalises “superstar” hubs (high variance = fragile to targeted attacks)
• Ollivier–Ricci curvature → rewards strong local clustering and shortcuts (positive curvature = more robust locally)

Result: a simple scalar that highlights structurally robust motifs even inside random graphs.

---

Quick start

# 1. Download the full package
wget https://github.com/syedrazaaftab/graph-robustness-score/raw/main/computational-testbed.zip

# 2. One-click reproduction
python notebooks/exploratory.py

How to choose the weights (w₁, w₂, w₃)
Default (recommended for most cases): w₁ = w₂ = w₃ = 1
This balances global connectivity, uniformity, and local clustering.
Quick rules of thumb:
•  Supply-chain / infrastructure graphs → increase w₁
•  Social / influencer networks → increase w₂
•  Biological / molecular graphs → increase w₃
The full package includes a weight-sensitivity heatmap.

### Install as Python package

```bash
pip install graph-robustness-score

from graph_robustness_score import compute_cg
import networkx as nx

G = nx.complete_graph(6)        # example
score = compute_cg(G)
print("C(G) =", score)

# Test it!
from graph_robustness_score import compute_cg
import networkx as nx

G = nx.complete_graph(6)  # example graph
score = compute_cg(G)
print("C(G) =", score)

**New in v0.1.0**: Full Ollivier–Ricci curvature is now available:

```python
# Fast version (default, same as before)
score = compute_cg(G)

# Full accurate Ollivier–Ricci (more precise, slightly slower)
score = compute_cg(G, use_full_ricci=True)

Install the full version with:

pip install "graph-robustness-score[full]"

See `examples/stock_example.py` for a real stock-sector demo (Energy sector scores highest right now).

Paper
“Complexity Functionals on Sparse Random Graphs: Spectral Gap, Singular-Value Variance, and Ollivier–Ricci Curvature Contributions”
Syed Raza Aftab, February 19, 2026
Exact analytic decomposition, regular-graph results, finite-size scaling up to N=800.

Commercial use & consulting
I offer paid network robustness audits (5–7 business days turnaround):
•  Supply-chain resilience reports
•  Fraud / transaction graph analysis
•  Molecular / protein interaction robustness
•  Cybersecurity attack-surface scoring
First 3 clients this month get 30% off + free 30-min call.
Email: [e-mail]
or book a free 15-min discovery call (Calendly link coming soon).

Made with ❤️ by Syed Raza Aftab
Independent Researcher, Princeton Meadows, New Jersey, USA
MIT License • Feel free to use the score in your own work (commercial or academic).
Just cite the paper or link back to this repo.