Impact Highlights
Improved MUTAG generalization by 7.5% via spectral regularization.
7.5Modeled traffic dynamics on a 100-node grid with spatio-temporal learning.
100Combined local ego-graph policies with deterministic fallback routing.
Mission Brief
Graph Neural Networks (Temporal and Spectral)
A set of graph-learning systems spanning ST-GCN forecasting, spectral regularization, and policy optimization with GraphSAGE.
PyTorch GeometricGraphSAGEGCNST-GCNSpectral Graph Theory
Build Notes
Developed an ST-GCN to forecast traffic over a 100-node grid from noisy time-series signals.
100Implemented spectral graph filtering with Laplacian regularization and improved MUTAG performance by 7.5%.
7.5Trained a GraphSAGE policy network for maze pathfinding using radius-2 ego-graphs and Dijkstra fallback.
2
Image Direction
Recommended Concept
A floating graph city where nodes pulse over time and edges animate as routing channels across a spatial map.
Text-to-Image Prompt
Futuristic network graph visualization, glowing nodes connected by animated edges, topographic data grid, emerald and blue glow, cinematic perspective, no text
Fallback asset in use: /illustrations/gnn-systems.svg