Huon Wilson

Hi @albvol. Thanks for filing an issue. I'd love to help, but I think I'll need some more information to be able to do this effectively: - some more code:...

Hi @AlexMRuch, thanks for filing an issue with such detail, and for trying to reverse engineer the algorithm to work around the poor error messages (if you're curious about the...

> The output above has 768 features for edges between submission-authors and submissions and 768 features for edges between comment-authors and submissions, and these are the main features I am...

I don't know of anyone actively working on this, sorry.

Hi @enjieli thanks for waiting! Unfortunately, I don't think HinSAGE is fully reproducible at the moment: https://github.com/stellargraph/stellargraph/pull/926 is a (very old) work-in-progress pull request that fixes this. In any case,...

You're right that large feature vectors may be suboptimal, because it does force copying a lot of data for every data point. HinSAGE samples a neighbourhood of each requested head...

Hi, we're keen to help. Could you say a bit more about your problem? Are you trying to take a collection of several graphs and predict attribute(s) for each one?...

Yeah, that should work. Depending on the problem there may be other approaches, for instance using a `exponential` activation, if the output value is only positive, or using a different...

Hi, thanks for waiting. > kept getting an error message saying some packages were incompatible, then nothing got installed It's been a while, but to help fix this I'll need...

Hi @fangxi, thanks for using StellarGraph and especially for filing an issue! I suspect the problem here is the version of `stellargraph` doesn't match the version of the notebook you...