EngineersofAI LogoCoursesBlogResearch LabAI LettersThe LabCode BankInteractive 3DKodrEarnestJobs
Skip to main content
Interactive 3D/GNN Message Passing
Aggregation
Message Passing
Feature Vectors
Round 0 features:
v0:
0.50
0.82
0.99
0.93
v1:
0.98
0.95
0.71
0.37
v2:
0.76
0.42
0.12
0.00
v3:
0.16
0.00
0.08
0.36
v4:
0.06
0.31
0.66
0.93
v5:
0.61
0.90
1.00
0.87
How It Works
Each round: every node collects its neighbors' features and averages them. After k rounds, each node sees its k-hop neighborhood.

GNN Message Passing - Interactive Visualization

Graph Neural Networks generalize convolution to graph-structured data. In each message passing round, each node aggregates feature vectors from its neighbors. After k rounds, each node's embedding captures its k-hop neighborhood. This visualization shows 3 rounds of message passing on a small graph, animating how information flows through the graph structure.

  • Watch node features aggregate from neighbors in each message passing round
  • See edges light up as messages flow between nodes
  • Choose aggregation: mean, max, or sum pooling
  • See feature vectors change after each round
  • Foundation for GCN, GraphSAGE, GAT, and molecular property prediction

Part of the EngineersOfAI Interactive 3D - free interactive visualizations covering every major concept in machine learning and AI engineering. Hover any element for a plain-English explanation. No code required.