Paper accepted at KDD 2020

We are happy to announce that our paper "Data Compression as a Comprehensive Framework for Graph Drawing and Representation Learning" has been accepted for presentation at the KDD 2020 conference, the top-level conference in data mining (ranked A*).

The paper by Claudia Plant, Sonja Biedermann and Christian Böhm presents an information-theroetic approach to graph embedding. The basic idea is to learn a low-dimensional vector space representation of a large graph by compressing the adjacency matrix. The low-dimensional coordiantes are optimized such that they support predicting the links in the graph with high accuracy. Experiments demonstrate that this novel approach yields natural representations of complex graphs which are suitable for clustering, link prediction and graph drawing.