3 ms·
Show HN: KarateClub a Python library for unsupervised machine learning on graphs
- chrisMyzel 7y agowould anyone please explain a practical non scientific example what one could do with this project given a arbitrary and complex graph being fed as an input
- np_tedious 7y agoI am by no means an expert in this space, but I'm pretty sure the most straightforward answers will be some form of "clustering". Find cliques among personal connections, business networks among companies, etc
- benitorosenberg 7y agoI am the OP. A few examples: 1. DeepWalk: You have the social network of users and for some of them their age. You want to predict the missing age values so that You are able to design a good targeted marketing campaign. The solution that DeepWalk gives: Learn a node level embedding, learn to predict the age based on the embedding, make a prediction for the missing users. 2. Graph2Vec: You have the molecules of hundred thousands of materials. You know that some of them is a carcinogen, while others are not. How can you predict whether a molecule is a carcinogen? The solution that Graph2Vec gives: Learn a graph level embedding, learn to predict the carcinogen / not carcinogen target and predict whether a molecule with unknown status is a carcinogen or not.
- Chris2048 7y agoThe tutorial link is fake
- benitorosenberg 7y agoIs dead. Dead does not mean fake.
- Chris2048 7y agoDead implies it was once a valid URL, but url is: https://github.com/benedekrozemberczki/karateclub/blob/master/link_here https://github.com/benedekrozemberczki/karateclub/blob/maste... the "link_here" text leads me to believe this is placeholder text in a fake URL that was intended to be replaced by a real one.