arXiv '20 | DyHATR: Modeling Dynamic Heterogeneous Network for Link Prediction using Hierarchical Attention with Temporal RNN

[jasperzhong]

https://arxiv.org/pdf/2004.01024.pdf

[jasperzhong]

也是DTDG，大概就是GATNE #341 for each snapshot，然后各个node embedding at each snapshot送到RNN里面得到一些时序信息，最后再做一个self-attention.

dgnn里面玩剩的老套路了.