Propagation Kernels
This provides a general solution for similarity measurement in various graph types, including labeled and attributed graphs, which is incremental as it builds on existing propagation schemes.
The paper tackles the problem of efficiently measuring similarity in structured data by introducing propagation kernels, a graph-kernel framework that uses information propagation schemes like random walks to capture structural information, resulting in faster computation than state-of-the-art approaches without sacrificing predictive performance.
We introduce propagation kernels, a general graph-kernel framework for efficiently measuring the similarity of structured data. Propagation kernels are based on monitoring how information spreads through a set of given graphs. They leverage early-stage distributions from propagation schemes such as random walks to capture structural information encoded in node labels, attributes, and edge information. This has two benefits. First, off-the-shelf propagation schemes can be used to naturally construct kernels for many graph types, including labeled, partially labeled, unlabeled, directed, and attributed graphs. Second, by leveraging existing efficient and informative propagation schemes, propagation kernels can be considerably faster than state-of-the-art approaches without sacrificing predictive performance. We will also show that if the graphs at hand have a regular structure, for instance when modeling image or video data, one can exploit this regularity to scale the kernel computation to large databases of graphs with thousands of nodes. We support our contributions by exhaustive experiments on a number of real-world graphs from a variety of application domains.