GNNs: An Introduction to Graph Neural Networks

Python 3.6+    |    Intermediate
  • 12 videos | 1h 21m 33s
  • Includes Assessment
  • Earns a Badge
Graph neural networks (GNNs) have recently become widely applied graph-analysis tools as they help capture indirect dependencies between data elements. Take this course to learn how to transform graph data for use in GNNs. Explore the use cases for machine learning in analyzing graph data and the challenges around modeling graphs for use in neural networks, including the use of adjacency matrices and node embeddings. Examine how a convolution function captures the properties of a node and those of its neighbors. While doing so explore normalization concepts, including symmetric normalization of adjacency matrices. Moving along, work with the Spektral Python library to model a graph dataset for application in a GNN. Finally, practice defining a convolution function for a GNN and examine how the resultant message propagation works. Upon completion you'll have a clear understanding of the need for and challenges around using graph data for machine learning and recognize the power of graph convolutional networks (GCNs).


  • discover the key concepts covered in this course
    outline graph data structures and common graph operations and describe the need for applying machine learning (ML) techniques on graph data
    recognize the need for node embeddings in setting up graphs for machine learning and describe how neural networks are constructed and are applied to graph data
    describe the different operations performed by individual neurons in a layer of a neural network
    outline graph convolutional networks (GCNs) and recognize the operations performed on input data when using a GCN, including symmetric normalization
    describe what knowledge graphs are and how graph neural networks (GNNs) can help uncover the information in such structures
  • set up the Python libraries required to use the Spektral library for building a graph neural network (GNN)
    define a graph structure which can be fed into a neural network using the Spektral library
    demonstrate the representation of nodes in a graph using a convolution function which accounts for its neighboring nodes
    illustrate the normalization of a graph's adjacency matrix based on the degree of its nodes
    demonstrate the factoring in of weights and the activation function for a neural network layer
    summarize the key concepts covered in this course


  • 2m 12s
  • 6m 37s
  • Locked
    3.  Graph Neural Networks (GNNs)
    9m 15s
  • Locked
    4.  A Neuron's Mathematical Operation
    7m 1s
  • Locked
    5.  Graph Convolutional Networks (GCNs)
    10m 19s
  • Locked
    6.  Knowledge Graphs - A GNN Use Case
    4m 14s
  • Locked
    7.  Installing Modules for Graph Neural Networks (GNNs)
    5m 49s
  • Locked
    8.  Creating Graph Structures Using Spektral
    10m 14s
  • Locked
    9.  Defining a Convolution Function for a GNN
    10m 41s
  • Locked
    10.  Building a Normalized Adjacency Matrix
    9m 57s
  • Locked
    11.  Using a Nonlinear Function in a Convolution
    3m 20s
  • Locked
    12.  Course Summary
    1m 53s


Skillsoft is providing you the opportunity to earn a digital badge upon successful completion on some of our courses, which can be shared on any social network or business platform.

Digital badges are yours to keep, forever.