search

lrjconan / GRAN

Efficient Graph Generation with Graph Recurrent Attention Networks, Deep Generative Model of Graphs, Graph Neural Networks, NeurIPS 2019
MIT License
462 stars 97 forks source link

Proper log_alpha normalization #8

Closed pclucas14 closed 4 years ago

pclucas14 commented 4 years ago

The previous version of the code was averaging alpha's, which sum to 1, in log space instead of in "regular" space. This resulted in alpha coefficients which did not sum to 1 anymore.

The following pull request merges the log alpha's before normalization, which fixes the problem.