how to deal with boundary problem about graph-learn

[eedalong]

Um, this question is not actually answered i think ==!

[hansugu]

From the code base, only hashed partitioning is supported

[jackonan]

Graph partition is a classical problem which was proposed many years ago. But till now, universal solutions are not found. In my point, we can develop kinds of task specific partition methods, and then push them to GL. These methods need to balance system efficiency and algorithm accuracy, and not only pursue the absolute correctness on system.

In our original paper, we propose some caching strategy to speed up sampling. It works well when graph data is huge. Due to the system complexity, it is not open-sourced in this version. Instead, we abstract Partitioner module, to enable solutions from contributors.

If you find some interesting solutions, welcome to contribution!