Temporal graph neural networks with Pytorch - How to create a simple recommendation engine on an Amazon dataset

Click for: original source

Over the course of the last few months, we at Memgraph have been working on something that we believe could be helpful with classical graph prediction tasks. With our latest newborn query module, you will have the option of performing both label classification and link prediction. By Antonio Filipovic.

The following is explained:

  • Graph neural networks
  • Temporal graph networks
  • Amazon data example
  • Exploring an Amazon data network in Memgraph

You probably already know that a graph consists of nodes (vertices) and edges (relationships). Every node can have its feature vector, which essentially describes that node with a vector of numbers. We can look at this feature vector as the representation vector of each node, also called embedding of the node. To avoid getting lost in technical details, graph neural networks work as a message passing system, where each node aggregates feature representations of its 1-hop neighbors.

Plenty of examples and code is provided as well in the article. Good read!

[Read More]

Tags cloud big-data devops data-science python