Dynamic Spatiotemporal GNN

0:00 / 0:00
Remixed 0 times

Created by userRonsil Felix Matthew Pacis with Mathify.dev

The animation visually walks through the Dynamic ST-GNN architecture, illustrating how input graph snapshots are processed through gated temporal convolutions, dynamic node embeddings, and adaptive adjacency matrices to produce multi-scale forecasts. It uses a clean, paper-like style with LaTeX equations and pastel-colored boxes to explain each component step-by-step.