fastgae

[wagpa]

https://github.com/deezer/fastgae

[wagpa]

Notes

• "Graph autoencoders (AE) and variational autoencoders (VAE) are powerful node embedding methods, but suffer from scalability issues"
• graph AE and VAE encodeing linear with number of edges per layer, decoding quadratic with number of nodes (+ memory issues for large n)
• Proposes sampling method by importance. It results in a biased approximate loss
• ...

[wagpa]

iterate over un-padded edges and calculate https://discord.com/channels/934839185855086662/988688735161946144/1070345614782632017

• does the padding change the result?
• weight edges higher than non-edges
• use as loss function and metric