mini-batches training for node classification on hypergraph

[Shuai Wang]

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
Hi, I am curious if it is possible to train hypergraph for node classification in mini-batches setting. Because now the graph data is growing bigger and bigger, training it in full-batch with hypergraph might not fit the GPU memory anymore. One direct way in my head is to chunk the incidence matrix but it may not work well if the message scheme is "node->hyperedge->node" like HGNN+? Hence, I want to ask if you have any suggestions on this problem.

Describe the solution you'd like
Some smarter ways to chunk the incidence matrix or some other methods?

Thanks in advance!