Can we convert the gaussian distribution to power law distribution?

[LynnZhang9]

https://github.com/rusty1s/pytorch_geometric/blob/a99b2cd19747c0287ff1fd3460e9cd99eaf84337/torch_geometric/nn/models/autoencoder.py#L219

Thank you for the great work! But I would like to know can we sample the real entities from another prior distribution for discriminator, like power law distribution? since normally the nodes in a social network(as an example:)), their basic features(in-degree, etc) fits power law distribution.

[rusty1s]

The Pyro package supports sampling from a variety of distributions, which might be helpful to you.