Graph Convolutional Networks (2016) (opens in new tab)

(tkipf.github.io)

168 pointsazthecx7y ago16 comments

16 comments

16 comments · 11 top-level

yazr7y ago· 2 in thread

There is an upcoming Manning ebook for graph-machine learning. Its a lot less technical than the papers. Fingers crossed for them.

https://www.manning.com/books/graph-powered-machine-learning

faizshah7y ago

Thanks for the recommendation, I haven't heard of this book yet!

Any recommendations on an up to date book on Social Network Analysis and Graph Clustering/Community Detection?

johnymontana7y ago

There is a new Graph Algorithms O'Reilly book that covers this area: https://neo4j.com/graph-algorithms-book/

ma2rten7y ago· 1 in thread

I didn't get a chance to look at the paper, but I wanted to share my experience working at two social network companies and publishing an academic paper on the topic of machine learning on a social network graph. Many people think it's necessary to use graph based machine learning (or graph databases).

In practice I found it almost always better to just use a "flat" representation and do regular machine learning (or database schemas). I would love to hear if others disagree.

wjn07y ago

I think it's highly task-specific. In network biology for example, taking advantage of the network structure, either through node embeddings or directly, is almost always helpful (even necessary). I'm not as familiar with applications to social networks, or graph DBs.

anonytrary7y ago· 1 in thread

> Also I'd like to thank Ferenc Huszar for highlighting some drawbacks of these kinds of models.

As someone who understands ML concepts but hasn't used ML in practice, can someone elaborate on these drawbacks compared to using a traditional neural network?

pfd19867y ago

In a traditional CNN you use the same convolution kernel across an entire image. Here you do the same but you convolve the features of a person (pixel) with the feature of its neighbors, but those change depending on the network. Drawbacks are: I) you add one new person to the network, the whole thing has to be retrained; ii) convolutions are more complex and expensive

jesseab7y ago· 1 in thread

Has there been any recent progress on GCNs or similar for embedding graphs rather than nodes?

wjn07y ago

Yes. This problem is of particular importance in drug response prediction, where small molecules are represented as graphs where the atoms are nodes and the bonds are edges. These are difficult to work with because of the variability in the number of nodes.

[1] https://arxiv.org/abs/1509.09292

[2] https://arxiv.org/abs/1603.00856

[3] http://proceedings.mlr.press/v48/niepert16.html

[4] https://persagen.com/files/misc/xu2017seq2seq.pdf (PDF)

stygiansonic7y ago

This related paper gives a great overview of ML methods on graphs: https://arxiv.org/abs/1709.05584

It was written by the authors of GraphSAGE, who also implemented it at Pinterest: https://arxiv.org/abs/1706.02216 and https://arxiv.org/abs/1806.01973

wjn07y ago

Variational graph auto-encoders, also by Kipf and Welling, is another great short, understandable paper. [1]

[1] https://arxiv.org/abs/1611.07308

plibither87y ago

Previous discussion (Show HN): https://news.ycombinator.com/item?id=12619694

gyre0077y ago

I recently came across this nice intro into how to get started with ML on Graphs [1]

https://medium.com/octavian-ai/how-to-get-started-with-machi...

xvilka7y ago

If somebody wants to help Julia community to support any types of GNN, there is an issue[1] for FluxML to support those.

[1] https://github.com/FluxML/Flux.jl/issues/625

JoeDaDude7y ago

A Keras-like tool for graph ML for those interested in tinkering: Warning: I have yet to useit. I am a complete noob in this area.

https://danielegrattarola.github.io/spektral/

apropostate7y ago

A great paper (http://openaccess.thecvf.com/content_cvpr_2017/papers/Monti_...) on Geometric Deep Learning, which builds upon the concepts used in GCNs.

j / k navigate · click thread line to collapse

16 comments

16 comments · 11 top-level

yazr7y ago· 2 in thread

There is an upcoming Manning ebook for graph-machine learning. Its a lot less technical than the papers. Fingers crossed for them.

https://www.manning.com/books/graph-powered-machine-learning

faizshah7y ago

Thanks for the recommendation, I haven't heard of this book yet!

Any recommendations on an up to date book on Social Network Analysis and Graph Clustering/Community Detection?

johnymontana7y ago

There is a new Graph Algorithms O'Reilly book that covers this area: https://neo4j.com/graph-algorithms-book/

ma2rten7y ago· 1 in thread

In practice I found it almost always better to just use a "flat" representation and do regular machine learning (or database schemas). I would love to hear if others disagree.

wjn07y ago

anonytrary7y ago· 1 in thread

> Also I'd like to thank Ferenc Huszar for highlighting some drawbacks of these kinds of models.

As someone who understands ML concepts but hasn't used ML in practice, can someone elaborate on these drawbacks compared to using a traditional neural network?

pfd19867y ago

jesseab7y ago· 1 in thread

Has there been any recent progress on GCNs or similar for embedding graphs rather than nodes?

wjn07y ago

[1] https://arxiv.org/abs/1509.09292

[2] https://arxiv.org/abs/1603.00856

[3] http://proceedings.mlr.press/v48/niepert16.html

[4] https://persagen.com/files/misc/xu2017seq2seq.pdf (PDF)

stygiansonic7y ago

This related paper gives a great overview of ML methods on graphs: https://arxiv.org/abs/1709.05584

It was written by the authors of GraphSAGE, who also implemented it at Pinterest: https://arxiv.org/abs/1706.02216 and https://arxiv.org/abs/1806.01973

wjn07y ago

Variational graph auto-encoders, also by Kipf and Welling, is another great short, understandable paper. [1]

[1] https://arxiv.org/abs/1611.07308

plibither87y ago

Previous discussion (Show HN): https://news.ycombinator.com/item?id=12619694

gyre0077y ago

I recently came across this nice intro into how to get started with ML on Graphs [1]

https://medium.com/octavian-ai/how-to-get-started-with-machi...

xvilka7y ago

If somebody wants to help Julia community to support any types of GNN, there is an issue[1] for FluxML to support those.

[1] https://github.com/FluxML/Flux.jl/issues/625

JoeDaDude7y ago

A Keras-like tool for graph ML for those interested in tinkering: Warning: I have yet to useit. I am a complete noob in this area.

https://danielegrattarola.github.io/spektral/

apropostate7y ago

A great paper (http://openaccess.thecvf.com/content_cvpr_2017/papers/Monti_...) on Geometric Deep Learning, which builds upon the concepts used in GCNs.

j / k navigate · click thread line to collapse