The aim of matrix completion is to find the values given a few of its entries. This is a topic of interest since Candes' article a few years ago. The aim of our project is to perform some matrix completion using a innovative graph-based approach. Graph theory is the current biggest topic in data science and machine learning as many social media and web communities provide large- scale graphs and one wants to be able to build recommender systems. We start describing the main principles of matrix completion on graphs, then we present a toy example simulating a synthetic netflix dataset in order to build a simplified recommender system.
A notebook and a report can be found in this repository