wintry ibex
Hello, I have been thinking about problem where I have pairs of graphs where each pair corresponds to an input graph that is transformed to an output graph by some transformation(s), e.g: an edge is added between two vertices, a new node is created with certain node attribute and connected to another vertex, a node attribute changes etc.
I am wondering whether it is possible to train a graph neural network to learn these transformations? I have tried looking through for papers and/or implementation on this but no luck yet. Thank you for your kind guidance!
polar obsidian