Machine Learning with Graphs

The work group Machine Learning with Graphs is headed by Ass.-Prof. Nils M. Kriege. It is part of the research group Data Mining and Machine Learning.


Graphs and networks are ubiquitous in various domains from chem- and bioinformatics to computer vision and social network analysis. Machine learning with graphs aims at exploiting the potential of the growing amount of structured data in all these areas to automate, accelerate and improve decision making. Analyzing graph data requires solving problems at the boundaries of machine learning, graph theory, and algorithmics.
The basic questions that arise such as deciding whether two graphs should be considered identical or quantifying their similarity pose both algorithmic and conceptual challenges.

We focus on the development of new machine learning and data mining methods for structured data related to the following broad topics:

  • Graph embedding
    These techniques generally map graphs into a vector space, such that similar graphs are represented by close vectors. We investigate graph kernels and graph neural networks for embedding graphs to solve learning problems.
  • Graph matching
    This term summarizes methods for finding a mapping between the nodes of two graphs that optimally preserves the adjacency structure according to a certain criterion. We study exact polynomial-time algorithms in restricted graph classes as well as heuristics for general large graphs.
  • Graph search
    We investigate efficient methods and index data structures for searching large databases of graphs, e.g., regarding similarity or subgraph containment.

Our ambition is to develop methods that are useful for solving concrete problems in real-world applications, especially in computational drug discovery.

Work Group