I am confused about the calculation order for the Sparse version

Open hbsun2113 opened this issue 7 years ago • 2 comments

hi, I read the code about the class SpGraphAttentionLayer, as following: In the original paper, I think operation2 is before the operation1. Can you tell me why you adjust the order?

Jan 29 '19 09:01 hbsun2113

@sh0416

Feb 02 '19 08:02 diegoantognini

The main difference between dense version and sparse version is softmax function.

There are no softmax function available in the sparse version, I calculate the nominator of softmax and the denomiator of softmax and then divide to get right softmax value.

There is a bug #16 which needs to be replace minus to substracting max value of each row.

Feb 02 '19 09:02 sh0416