liveBook · Manning

End-to-end model

As explained in Graph Neural Networks in Action

A model that jointly learns intermediate representations and task predictions during a single training process.

An end-to-end GNN jointly learns node embeddings and label predictions in one training process, rather than passing static embeddings into a separate classifier. section 2.3.3

A practical implementation follows this pipeline:

  • Prepare the graph for PyTorch Geometric and compare randomized node features with N2V embeddings. section 2.3.3
  • Modify a SimpleGNN with GCN layers, log softmax output, and dual outputs for embeddings and predictions. section 2.3.3

Books graph classification

On the books_gml dataset, the four-layer GNN with N2V features reaches 88.99% accuracy and 89.29% F1, compared with lower results for randomized features.

How Graph Neural Networks in Action covers it

How deep it goes
Explains data preparation, architecture changes, masked optimization, and embedding inspection, but not deeper training theory.
Examples
Compares two- and four-layer GNNs using randomized or N2V node features on the books_gml graph.
What it assumes
Assumes a graph dataset and a SimpleGNN workflow, while demonstrating how to prepare features and outputs for classification.
What you can do after
Build it from scratch with PyTorch Geometric, train masked node classifiers, and inspect learned embeddings alongside predictions.
How it connects
Connects graph neural networks, graph embeddings, Node2Vec, node classification, and random forest models.
Cover of Graph Neural Networks in Action
liveBookmanning.com
Read Graph Neural Networks in Action on liveBook Get the book at manning.com

Also explained in

Compare how 2 books explain end-to-end model →

This explainer was generated by Manning's knowledge graph from the text of Graph Neural Networks in Action. The section links point to the parts of the book it draws on.