Temporal Node Features

[ivankrstev7]

Hi, I’m doing experiments on financial data using GraphMixer and other temporal graph networks, and I’ve found that GraphMixer consistently outperforms the others by a significant margin.

That said, I have a question regarding the node features. In my current setup, the node features are static over time, which isn’t ideal for financial data where temporal dynamics are crucial. I’m wondering if there’s an efficient way to incorporate time-varying node features.

As a workaround, I’ve been using edge features to represent the changing attributes of nodes, but I’m confident that directly applying temporal node features could lead to even better performance. Any guidance on how to implement this would be greatly appreciated.

Thank you!