search

snap-stanford / ogb

Benchmark datasets, data loaders, and evaluators for graph machine learning
https://ogb.stanford.edu
MIT License
1.93k stars 397 forks source link

Why Pre-compute GCN normalization is not used on ogbn-arxiv datasets? #130

Closed nbl97 closed 3 years ago

nbl97 commented 3 years ago

In the folder nodeproppred, all the adj_t have been pre-computed except ogbn-arxiv when selected model is GCN. The default GCNConv can do the same thing, why do we pre-compute? Thanks a lot.

rusty1s commented 3 years ago

We pre-compute the normalization coefficients of GCN to be a bit more efficient w.r.t. to runtime and memory, which is especially useful on those larger datasets. You can do the same for ogbn-arxiv as well, although it is not necessary to fit the GNN on the GPU.