StellarGraph

Recommended Algorithms for a Tree Graph - feasibility of intermediary nodes

Hi;
Since my knowledge about graphs is limited I’d like to know what do you recommend for node classification in a tree graph?
Does intermediary nodes which doesn’t have any features and just indicate they are parents of some nodes have any advantage in these algorithms?

Thank you for providing this framework and community.

Hi,

Thanks for your interest in StellarGraph! In general, StellarGraph does not have any special algorithms for tree graphs. Also, nodes that do not have features will not be useful for the current algorithms in StellarGraph, so I would suggest modifying graphs to remove intermediate nodes and only connect nodes with features. Could you tell us more about the graph, such as what the nodes represent, which nodes have features, and which nodes you would like to classify?

You might also be interested in an alternative algorithm that is specifically for tree graphs: https://github.com/stanfordnlp/treelstm

Regards,
Andrew

1 Like

My graphs are created from webpages, loosely representing HTML DOM elements. Their features are derived from DOM elements. The problem is a true/false classification of all nodes.
I’m just looking for any algorithm that will work on this kind of graphs, not necessarily a special algorithm.
Thank you for your help

It sounds like you may be able to use StellarGraph for your problem, the node classification with GCN demo would be a good place to start. Please let us know how things work out for you!

Regards,
Andrew