benitorosenberg

Participated in their competition and won $50,000 - developed an ML system based on their DB and I can tell that their product is completely broken. Does not scale, poorly documented. The whole company felt like a marketing stunt. Would be surprised if they would be around in the future.

Sounds like a good idea, might generate a dynamic visualization on some examples. :)

The Leskovec papers in the readme are pretty good (classical pieces) I would really recommend reading them.

TY! I fixed it.

Entire graph embedding means that you have a lot of smaller graphs (e.g. molecules, transactions, threads) and you want to classify them. We created this package which covers these methods:

https://github.com/benedekrozemberczki/karateclub

The reason for not doing that is the bias that such sampling introduces.

We are writing a paper out of this, but the main point is that you can achieve these two things with minimal classification performance degradation:

1. Speeding up node embedding and classification. 2. Speeding up whole graph embedding and classification.