Graph Neural Network Lab
DLArabic GNN Lab
Learn graph neural networks hands-on: draw a graph, follow message passing between nodes step by step, really train GCN, GraphSAGE, GAT and GIN in your browser, predict links, explore representations, and understand why the model made its decision. Every result is labelled with its kind: mathematical illustration, simulation, real training or inference.
Learn graph neural networks
Detailed academic explanations that complement the interactive modules: read the lesson, then open the model in the guide, then try it in the lab.
Interactive modules
Each module combines an explanation and equations with an experiment that really runs in your browser, without sending your data to any server.
Build and understand
From a graph you draw by hand to the equation that drives each layer
Train and evaluate
Real training in the browser with a split and honest measurement of results
Understand the decision
Why the model predicted what it predicted
From the lab to research
Documented data, applications, projects and a quiz
What do the labels next to each result mean?
The lab never shows a number without saying how it was computed. Distinguishing the four kinds is part of the scientific method here.
A deterministic mathematical computation on small values that you can see in full (no learning involved).
A simulation that explains the mechanism with simplified values or fixed weights; it is not a trained model.
Real training with gradients and backpropagation in your browser, on small real data or declared synthetic data.
Running a model that was actually trained on new inputs and reading its outputs.
