Message passing

As explained in Knowledge Graphs and LLMs in Action

Message passing is a graph-neural-network technique in which nodes iteratively exchange, aggregate, and use information from neighboring nodes to update their representations.

Message passing lets graph neural networks update each node by collecting information from neighbors, processing it, and combining it with the node’s representation in repeated rounds. section 11.5.1

A typical message-passing pipeline proceeds as follows:

• Collect messages from neighboring nodes. section 11.5.1
• Aggregate the messages, commonly by summation, then update the node representation with trainable weights, a nonlinear activation, and optionally a bias. section 11.5.3
• Repeat the process so that k iterations expose each node to information from its k-hop neighborhood. section 11.5.2

Social-network interests

A user representation can begin with explicitly stated interests, incorporate friends’ interests in the first round, and incorporate friends-of-friends’ interests in the second. Later rounds capture broader social patterns. section 11.5.1

Adding self-loops folds a node’s own information into neighborhood aggregation, simplifying implementation and often improving stability, but it reduces flexibility in distinguishing self-features from neighbor features. Specialized aggregation, attention, skip connections, and normalization extend the framework for irregular neighborhoods and deeper models. section 11.5.4 section 11.6

Figure 11.5 Example of how a single node aggregates messages from its local neighborhood in the message-passing model. Notice that the computation graph of the GNN forms a tree structure by unfolding the neighborhood around the target node.

Unfolds a target node’s local neighborhood into a computation tree, showing how messages are aggregated during graph-neural-network inference. section 11.5.1

How Knowledge Graphs and LLMs in Action covers it

How deep it goes

Defines aggregation and update, gives a basic trainable equation, and compares self-loops with specialized architectural enhancements.

Examples

Uses social networks, molecular graphs, citation networks, and an unfolded neighborhood computation tree.

What it assumes

Assumes graph nodes, edges, neighbors, and node features while introducing iterative aggregation and representation updates.

What you can do after

Build a basic GNN update pipeline and choose between standard aggregation and self-loop variants.

How it connects

Connects graph neural networks with graph topology, attention mechanisms, skip connections, normalization layers, and LLMs.

liveBookmanning.com

Read Knowledge Graphs and LLMs in Action on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Graph Neural Networks in ActionMessage passing is a graph neural network technique in which node representations are updated by aggregating information from neighboring nodes.
• 

  liveBookmanning.com
  Optimization Algorithms: AI techniques for design, planning, and control problemsMessage passing is a graph-learning technique in which nodes exchange and aggregate information with neighboring nodes to update representations.
• 

  liveBookmanning.com
  Learn Concurrent Programming with GoMessage passing is a communication approach in which concurrent executions exchange information by sending and receiving messages.
• 

  liveBookmanning.com
  Elixir in Action, Third EditionMessage passing is interprocess communication in which processes exchange data through runtime messages rather than shared memory.
• 

  liveBookmanning.com
  Grokking ConcurrencyMessage passing is an interprocess communication technique in which execution contexts exchange data through discrete messages rather than directly sharing memory.
• 

  liveBookmanning.com
  C++ Concurrency in Action, Second EditionMessage passing is a communication technique in which concurrent components exchange information by sending and receiving messages.
• 

  liveBookmanning.com
  Sutskever's ListMessage passing is an iterative graph-computation technique in which nodes send information to neighboring nodes and update their representations.
• 

  liveBookmanning.com
  Machine Learning Algorithms in DepthMessage passing is a graph-computation technique in which nodes exchange information with neighboring nodes to update their representations.

Compare how 9 books explain message passing →

Related concepts

• shared memoryalternative to · Shared memory is a memory region or storage area accessible to multiple processes, threads, or other execution components for accessing or exchanging data.
• actorused by · An actor is an independent entity or participant that performs actions, maintains or pursues state or goals, and interacts with other entities or systems.
• distributed systemused by · A distributed system is a computing system whose components run across multiple networked processes or machines and coordinate through communication.
• graph neural networkused by · A graph neural network is a neural-network model that learns representations or predictions from graph-structured data by combining information from connected entities.
• embeddingproduces · A numerical vector representation of an item or input that encodes features or relationships useful for computational tasks.
• communicating sequential processused by · Communicating Sequential Processes is a formal concurrency model in which independent sequential processes communicate through channels by exchanging messages.

This explainer was generated by Manning's knowledge graph from the text of Knowledge Graphs and LLMs in Action. The section links point to the parts of the book it draws on.