Finding K-Hop Neighbors

[YoavLotem]

Hey :)

I'm wondering if its possible to utilize the (excellent) PyG library in order to find k-hop neighbors in a large scale graph database.

How does the PyG "neighbor finding" engine works and where can I find the code?

So lets say for example I have a large sparse/dense adjacency matrix and I want to multiply it by itself in order to find 2-hop neighbors of each vertex. Does PyG has an optimized method in doing so or do it have a smarter way of finding neighbors?

Thanks!

[LukeLIN-web]

(https://pytorch-geometric.readthedocs.io/en/latest/_modules/torch_geometric/utils/subgraph.html)
https://pytorch-geometric.readthedocs.io/en/latest/modules/utils.html#torch_geometric.utils.k_hop_subgraph

[rusty1s]

Alternatively, you can use NeighborLoader with num_neighbors=[-1] * k

[ChanglongZheng]

I have this question too, trying k_hop_subgraph and finding it helpful.