Graph-Conditioned Mixture of Graph Neural Network Experts for Traffic Forecasting

In three linesGC-MoE is a graph-conditioned mixture-of-experts framework for spatio-temporal traffic forecasting. It assigns each node a personalized combination of frozen GNN experts based on graph topology and recent input windows. Trained on PEMS04, PEMS07, METR-LA, and PEMS-BAY, the model improves MAE with only ~17K trainable parameters on top of 1.5M frozen expert weights.

Summary generated by Claude — human-verified