How to use sample_from_edges

[LuisaWerner]

Hi,

my goal is to sample batches from a Data object based on the edge index.
I have a graph with node features [n x d] (in data.x) and edge index [2 x num_edges] (in data.edge_index), I want to create batches based on the edges index so that each batch keeps the original features data.x as in the original object but a subset of the edges.

The edges should be sampled subsequently. If the batch-size was 10, the first batch should have the first 10 batches in edge_index [2 x 10], the second batch the next 10 edges and so on.

Link_Neighbor_Loader does not do what I want to achieve since I don't need any neighbors but the subsequent sampling of edges.
Edge_Sampler_Input looks like being close to what I want to do but I don't know how to use sample_from_edges or how to overwrite it.

Can someone help me with this problem?
Thanks in advance!
luisa

[wsad1]

pyg has a RandomNodeLoader. You probably want a RandomeEdgeLoader. This can be built by modifying the RanodomNodeLoader code by

1. while calling super().__init__ change self.num_nodes to self.num_edges
2. in collate_fn call self.data.edge_subgraph.

Let me know if you need help making these updates to RandomNodeLoader.