search

benedekrozemberczki / karateclub

Karate Club: An API Oriented Open-source Python Framework for Unsupervised Learning on Graphs (CIKM 2020)
https://karateclub.readthedocs.io
GNU General Public License v3.0
2.17k stars 247 forks source link

evaluate and compare different graph embeddings #94

Closed SalvatoreRa closed 2 years ago

SalvatoreRa commented 2 years ago

Hi,

I have check the documentation and th repository but I have not found evaluation metrics to evaluate the quality of the embeddings. I have seen that embedding can be used for classification and thus running on it a classifier is a way to check if the embedding is informative or not. however, not always you have ground truth value and would be it better to have some quality metrics.

I have seen in different articles that they are using Man Average Precision or reconstruction error to evaluate the quality of embeddings.

There are metrics implemented or planned to be implemented?

benedekrozemberczki commented 2 years ago

You can formulate a link prediction problem every time if that makes sense. Do you know what I mean @SalvatoreRa ? Can you star the repo?