# Graph Convolutional Matrix Completion

**Authors:** Rianne van den Berg, Thomas N. Kipf, Max Welling  
**arXiv:** [1706.02263](https://arxiv.org/abs/1706.02263)  
**Date:** 2017-06-07  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1706.02263v2/

## Key points

- The paper proposes a novel approach to matrix completion for recommender systems using graph-based techniques.
- Matrix completion is viewed as link prediction on graphs, where interaction data can be represented by a bipartite user-item graph with labeled edges denoting observed ratings.
- The proposed approach builds upon recent progress in deep learning on graph-structured data and introduces a graph auto-encoder framework based on differentiable message passing on the bipartite interaction graph.
- This framework shows competitive performance on standard collaborative filtering benchmarks.
- The authors demonstrate that their approach outperforms recent state-of-the-art methods in settings where complimentary feature information or structured data such as a social network is available.
- The proposed method leverages this additional information to improve its predictions.
- Overall, the paper presents an innovative and effective solution to the problem of matrix completion for recommender systems using graph-based techniques with significant potential for real-world applications in various domains including e-commerce and entertainment.

## Summary

The paper titled "Graph Convolutional Matrix Completion" by Rianne van den Berg, Thomas N. Kipf, and Max Welling proposes a novel approach to matrix completion for recommender systems using graph-based techniques. The authors view the problem of matrix completion as link prediction on graphs, where interaction data such as movie ratings can be represented by a bipartite user-item graph with labeled edges denoting observed ratings. The proposed approach builds upon recent progress in deep learning on graph-structured data and introduces a graph auto-encoder framework based on differentiable message passing on the bipartite interaction graph. This framework shows competitive performance on standard collaborative filtering benchmarks. Moreover, the authors demonstrate that their approach outperforms recent state-of-the-art methods in settings where complimentary feature information or structured data such as a social network is available. The proposed method leverages this additional information to improve its predictions. Overall, the paper presents an innovative and effective solution to the problem of matrix completion for recommender systems using graph-based techniques with significant potential for real-world applications in various domains including e-commerce and entertainment.