Dynamic graph neural network

As explained in Graph Neural Networks in Action

A graph neural network model that represents and predicts changes in graph structure or node and edge information over time.

A dynamic graph neural network models relational data as it changes over time. It connects sequential graph snapshots so recent observations can support predictions about unseen future events. section 6.0 section 6.3

Use a dynamic GNN when both relationships and their associated information evolve:

• Model changing friendships, road conditions, traffic patterns, or arrival-time predictions. section 6.1
• Predict trajectories of interacting bodies, including molecules, cells, objects, or stars. section 6.1
• Represent temporal data by combining graph snapshots and adding temporal edges between corresponding nodes. section 6.3

Pose estimation

Treat a walking body as a relational graph observed across video frames. A model learns from earlier poses and predicts the next step, either with a temporal graph and masked nodes or with snapshot encoders followed by VAEs and RNNs. section 6.0 section 6.3

Figure 6.1 Examples of different dynamic problems

Compare dynamic problems involving social relationships, traffic, and interacting trajectories. section 6.1

How Graph Neural Networks in Action covers it

How deep it goes

Builds temporal graphs with snapshot connections, masking, GATs, VAEs, and RNNs, while deferring further technical details.

Examples

Uses changing friendships, road traffic, interacting trajectories, and walking-pose video frames as concrete dynamic-graph problems.

What it assumes

Builds on previously introduced graph-learning concepts and assumes familiarity with the surrounding foundational material.

What you can do after

Build it from scratch to connect graph snapshots and predict future graph evolution from recent observations.

How it connects

Connects graph neural networks with temporal learning, graph attention networks, recurrent neural networks, variational autoencoders, and neural relational inference.

liveBookmanning.com

Read Graph Neural Networks in Action on liveBook Get the book at manning.com

Related concepts

• graph attention networkuses · A graph attention network is a graph neural network that uses attention weights to aggregate information from neighboring nodes, allowing their contributions to differ.
• neural networkkind of · A neural network is a computational model of interconnected, parameterized units that learns transformations from inputs to outputs.
• pose estimationaddresses · The task of estimating the configuration or movement of body parts from observed data.

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.