Graph Neural Network

• Evey node of the graph has some kind of message attached to it
• Messages can be passed to other nodes in the graph they are adjacent to

• The reciever will process all incomming messages with some symmetric function (for exaple their sum , , or )

  • Order independent
  • Number-of-children independent
  • The output of the node takes into account it’s own message and the aggregated children messages