For blogs, I'd recommend: - @thomaskipf's post on Graph Convolutional Networks: https://tkipf.github.io/graph-... - My blog on Graph Attention Networks: https://petar-v.com/GAT/ - A series of comprehensive deep-dives from @mmbronstein: https://towardsdatascience.com...

For a comprehensive overview of the area in the form of a talk, I would highly recommend @xbresson's guest lecture at NYU's Deep Learning course: https://x.com/alfcnz/status/13...

For keeping up with the latest trends in graph representation learning, @SergeyI49013776 maintains a very useful Telegram feed: http://ttttt.me/graphML, as well as a recently-launched GRL newsletter: http://newsletter.ivanovml.com...

For access to the most recent strong GRL benchmark datasets, I would recommend the OGB (http://ogb.stanford.edu) by @weihua916 et al., and Benchmarking-GNNs: https://github.com/graphdeeple... by @vijaypradwi, @chaitjo et al.

For quickly getting started with GRL implementations, check out PyTorch Geometric by @rusty1s: https://github.com/rusty1s/pyt... and DGL by @GraphDeep: https://www.dgl.ai/ For a repository containing the most curated set of GRL papers, tutorials etc, check out: https://github.com/naganandy/g...

For an awesome over-arching textbook resource on the entire field, consult the recent GRL book by @williamleif: https://www.cs.mcgill.ca/~wlh/... For excellent university courses, check out CS224W by @jure: https://web.stanford.edu/class... and COMP 766 by @williamleif: https://cs.mcgill.ca/~wlh/comp...

Any further resources I might have missed? Feel free to comment at any part of this thread. Hope you'll find it useful! 😊

@PetarV_93 Thanks for the shout-out! I would also add the Graph Networks survey by Battaglia etal, DeepMind, for a thought-provoking take on GNNs and inductive biases. I still find myself re-visiting it.

@chaitjo No worries, it's more than well-deserved. Thanks for chiming in! :) I deliberately decided to avoid (survey) papers from this thread, because there's so many of them and each explores a different but exciting angle. That being said, of course the GNs paper is super insightful.

@PetarV_93 Thanks for sharing these! Are GNNs suitable for information retrieval from webpages?

@ajitk I would say so! There's some recent examples of GNNs being quite useful for mining Wikipedia, for example. See #WikiCS from @pmernyei1 for one example: https://arxiv.org/abs/2007.029...

@PetarV_93 I remember 4 years ago learning GNNs/GCNs from the first @thomaskipf papers and yours as well. It is fascinating to see now there is so much material also from University courses! I am also glad to be part of the community that made it possible!

@PetarV_93 Penn is currently running a course on GNN. Lectures and course material are available here: https://gnn.seas.upenn.edu/

@PetarV_93 This is a great list of resources Petar! I also recommend this resource, I use it mostly for libraries (for GNNs and other subjects) - https://madewithml.com/topics/...

@PetarV_93 Ever wondered,why do GNNs fail to perform better when we add more layers. Cz, adding more layers should have definitely helped to better model complex relationships. https://link.medium.com/UWznPm...

@PetarV_93 For a moment I thought these were higher order Feynman diagrams

@PetarV_93 This is my favorite GNN intro: https://distill.pub/2021/gnn-i...

@PetarV_93 extremely interesting

@PetarV_93 Just what I was looking for, thanks!

@PetarV_93 Cool

@PetarV_93 @threadreaderapp unroll

@PetarV_93 @threadreaderapp compile

@PetarV_93 @threadreaderapp unroll

@PetarV_93 @shakir_za @threader_app compile

@PetarV_93 Thanks very much for this

@PetarV_93 @threadreaderapp unroll

@PetarV_93 @threadreaderapp unroll