Impact of Node Features on GNNs for Node Classification

[fafal-abnir]

Hi everyone,

I'm currently exploring the impact of various node features on the performance of Graph Neural Networks (GNNs) for node classification tasks. I've come across features like PageRank, HITS, and community-based attributes that seem to enhance classification accuracy by providing additional contextual information.

I would love to hear your thoughts on the following:

1. How do you integrate features like PageRank or HITS into your GNN models, and what impact have you observed on node classification performance?
2. Are there specific methods or frameworks that you recommend for effectively incorporating community-based features into GNNs?
3. Have you encountered any research papers or resources that provide insights or case studies on this topic?

Thanks in advance for your insights!