Any Support for Heterogeneous Graph?

Questions
#1

I want to use dgl for a Heterogeneous graph, Is there any examples?

#2

Heterogeneous graph support will be in our 0.4 release.

What models do you like to use on heterogeneous graphs? Having more models for heterogeneous graphs will be a great help for our API design.

Thanks

#3

Thanks for your reply, it seems metapath is the most common way in HIN, I found the following paper to be helpful.

Wang, X., Ji, H., Shi, C., Wang, B., Cui, P., Yu, P., & Ye, Y. (2019). Heterogeneous Graph Attention Network. CoRR, abs/1903.07293 .

#4

Thanks for your suggestions. metapath seems a promising solution. We’re actually looking into metapath2vec. We’ll also look into the paper you suggested and try a few options. We’ll keep you updated.

#5

Hi, I found that v0.3 just release, is there any roadmap for v0.4? And I found a branch is for heterogeneous graph, and document about the is available? Thanks!

#6

Following is the PR for Heterograph’s API.


The roadmap for v0.4 is not ready yet, please stay tuned!

#7

Here you go: https://github.com/dmlc/dgl/issues/666 . Heterograph will be the major feature in the next release.

1 Like
#8

Heterograph will be a great release version. May I ask that how long will it be ready for an initial release？

#9

Our typical release cycle is two to three months.

1 Like
#10

Can you please say when we can expect DGL to roll out support for a knowledge graph? I would need it for my work on developing a recommendation system based on a knowledge graph.