liveBook · Manning

Multi-aggregation

As explained in Graph Neural Networks in Action

Multi-aggregation is a graph neural network technique that combines multiple aggregation functions within one layer or across layers.

Multi-aggregation combines neighborhood summaries to expose different graph signals. Use it either across layers—such as mean followed by max—or within one layer using several aggregators. section 3.2.2

In PyTorch Geometric, the main patterns are:

  • Assign each GraphSAGE layer its own function, such as mean for smoothing followed by max for emphasizing prominent neighbor features. section 3.2.2
  • Pass a list such as ['mean', 'max', 'sum'] to SAGEConv, or provide aggregation module instances, so one layer captures several neighborhood properties. section 3.2.2

How Graph Neural Networks in Action covers it

How deep it goes
Explains layered choices, list syntax, module instances, and comparative metrics, but stops short of aggregation internals.
Examples
Uses mean, max, and sum combinations, SAGEConv declarations, and GraphSAGE classification tables.
What it assumes
Assumes GraphSAGE layers and neighborhood aggregation, while demonstrating the PyG configuration from concrete examples.
What you can do after
Apply it via PyTorch Geometric by configuring SAGEConv with layered functions, string lists, or aggregation modules.
How it connects
Connects GraphSAGE, neighborhood aggregation, PyTorch Geometric, MessagePassing, and Jumping Knowledge.
Cover of Graph Neural Networks in Action
liveBookmanning.com
Read Graph Neural Networks in Action on liveBook Get the book at manning.com

Related concepts

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.