A MATLAB implementation of the propagation graph kernel for general graphs (propagation_kernel.m) and grid graphs (propagation_kernel_grid.m). The propagation kernel is described in:
Marion Neumann, Roman Garnett, Christian Bauckhage, Kristian Kersting. Propagation kernels: efficient graph kernels from propagated information. (2015). Machine Learning. 102 (2), pp. 209-245.
and
Neumann, M. Patricia, N., Garnett, R., and Kersting, K. Efficient Graph Kernels by Randomization. (2013). Machine Learning and Knowledge Discovery in Databases: European Conference (ECML/PKDD 2012), pp. 378--392.
This implementation supports attributed, labeled and weighted graphs and arbitrary user-defined base kernels and transformations as well as arbitrary user-defined neighborhoods for the grid version.
Add the directory to your MATLAB path and use help propagation_kernel to view the documentation. A simple demo run_default_PK.m
in the demo/ directory shows how the setup the parameters and run the kernel computation with defaults for labeled graphs. Further demos show how to run the kernel with other than default settings for partially labeled graphs
(demo.m), attributed graphs (demo_p2k.m)
and grid graphs, i.e., graphs of regular neighborhood structure, (demo_grid.m).