Stanford CS224W: ML with Graphs | 2021 | Lecture 13.4 – Detecting Overlapping Communities

For more information about Stanford’s Artificial Intelligence professional and graduate programs, visit: https://stanford.io/3GsVmnL

Lecture 13.4 – Detecting Overlapping Communities BigCLAM
Jure Leskovec
Computer Science, PhD

We introduce how to find overlapping communities in a graph. For example, detecting social groups in a Facebook friendship network. We will define Community Affiliation Graph Model (AGM) which is a generative model for graphs based on community affiliations. We see how AGM randomly assigns two nodes to the same community, and how can we generate a whole graph using it. Then given a graph, we show how to find the best AGM in order to find the communities. We finally show how to maximize likelihoods and find the parameters of the best model which will lead to the BigCLAM model and we explain its working.

To follow along with the course schedule and syllabus, visit:
http://web.stanford.edu/class/cs224w/

#machinelearning #machinelearningcourse