Ricci Flow KI: Graph Optimization System for GNNs

A production-ready system implementing Ricci flow-based graph rewiring to improve Graph Neural Network (GNN) performance on heterophilic graphs. Achieves +13% average accuracy improvement through discrete Ricci curvature optimization (SDRF, BORF, Adaptive methods) with hyperparameter tuning via Optuna. Features: - Three rewiring algorithms (SDRF, BORF, Adaptive) - GPU-accelerated training - Hyperparameter optimization - Comprehensive benchmarks on Texas, Cornell, Wisconsin datasets - Production-ready Python implementation Results: 67.80% accuracy on Texas dataset (+15.37% over baseline)