This is a tutorial for computing various kernel matrices using the graphkernels package. An example script is given in demo_mutag.py.

The file demo_mutag.py can be executed as

python demo_mutag.py

The script will compute all the available kernels on the benchmark MUTAG dataset.

The data that the graph kernels functions require as input, should be a list of igraph objects, as provided in the file data.mutag. This example file is available here in the tutorial, but is also included in the downloadable package.

Threee main steps are required in Python to compute a kernel matrix.

1. Import the packages

    import numpy as np

    import graphkernels.kernels as gk

2. Load the data

   mutag_list = np.load('data.mutag')

3. Compute the kernels: example with the WL kernels

   K_wl = gk.CalculateWLKernel(mutag_list, par = 3)

The matrix K_wl is the kernel matrix, obtained with the WL kernel, and therefore is a square matrix of size equal to the number of samples. The par here represents the number of iterations of the WL kernels. Note that if no parameters is provided, a default value is used.

Here is a complete list of all the graph kernels that can be computed with our package, the python functions to use and corresponding kernels parameters.