e205ee2a5de471a70c1fd1b46033a75f-AuthorFeedback.pdf

Wevalidated this guess bychecking the variance explained, which are 59% (left)8 and 17% (right). This also indicates that our model learns amore compact latent space. Thank you for raising this9 question. For example, one can make itO(N) by15 predicting all edges of a node at the same time. Thus, the uniqueness is higher but the reconstruction accuracy is lower (Table 1).