The Knowledge Graph Conference Icon

The Knowledge Graph Conference

Home
Events
Members

Announcements
Ask
Ask The Ontologists
Events
Jobs
Promotions
Share

Powered by Tightknit

GNNs on Knowledge Graphs: Use Cases and Tips for Success

·Jun 09, 2021 01:06 PM

Hi everyone. Anybody used GNNs on a KG? Any good use cases anyone can share? Examples? Things to expect/avoid? I’m planning to, but going into it blindly, I don’t know what to expect.

7 comments

· Sorted by Oldest

Ellie Y.
·
I believe Louis G. has experience here, and Edward E. might know of some good references for getting started
🙌1
Edward E.
·
Lalla Ellie Y. this is a good SOTA paper https://arxiv.org/pdf/1911.06962.pdf ! Also for an overview of general ml approaches for KG reasoning (i.e. multi-hop reasoning instead of just completion, this is a good survey paper https://www.sciencedirect.com/science/article/abs/pii/S0957417419306669)
🙌2
Ellie Y.
·
i knew it! 😄 thanks Edward!
🔥1
➕1
Diego Esteves
·
We benchmarked different SOTA models but what I can share for now is that results aren't yet that good for real use cases. Maybe better to understand what do you want to accomplish before deep diving into GNNs? At this moment, I'd not bet all my chips on it Lalla...but generally speaking topic is quite promising
👍1
Lalla
·
Got it, got it, and got it! Thank you all for your responses!
Nariman A.
·
Checkout the tutorial given by Daniel V. at the Graph-based data science workshop. He used kglab and rubrix for semisupervised classification of recipes based on relations with ingredients. Both Kglab and rubrix are open source with examples on github.
🙌1
Daniel V.
·
Thanks so much for mentioning Nariman A.! Lalla here's the tutorial from the Graph-based DS workshop, using kglab, pytorch geometric and Rubrix: https://docs.rubrix.ml/en/stable/tutorials/03-kglab_pytorch_geometric.html#%F0%9F%A7%AA-Node-classification-with-kglab-and-PyTorch-Geometric
👍2