graphkernels-review: Code and data sets for the review on graph kernels

This repository contains the scripts used to run the experiments seen in the following article:

Karsten Borgwardt, Elisabetta Ghisu, Felipe Llinares-López, Leslie O'Bray and Bastian Rieck (2020). Graph Kernels: State-of-the-Art and Future Challenges, Foundations and Trends® in Machine Learning: Vol. 13: No. 5-6, pp 531-712.

Using the repository

As a general structure, we assume to have the data stored in a data folder and the generated kernel matrices saved in an output matrices folder. We used poetry to manage the dependencies, which can be installed by running poetry install in the terminal (assuming poetry is already installed).

Generating kernel matrices

./src/create_kernel_matrices.py -o ./matrices/MUTAG ./data/MUTAG/*.pickle

Training a classifier on a set of kernel matrices

./src/train.py ./matrices/MUTAG/*.npz -n MUTAG -o ./results/MUTAG.json

Example illustration of what output looks like

The output is stored in a json file with results, and can be aggregated into a single csv file using the script:

./src/analyse_multiple.py ./results/*.json

	0	1	2	3	4	5	6	7	8	9	mean	std
EH	85.51	85.01	84.54	85.17	84.48	85.73	84.59	85.78	84.62	83.95	84.94	0.61
GL	67.63	68.10	66.52	67.57	66.52	68.74	67.07	67.54	69.82	67.07	67.66	1.02
SP	83.12	82.98	79.82	84.48	84.65	83.18	83.46	82.09	84.56	85.50	83.38	1.61
VEH	84.57	82.42	81.84	82.12	85.15	86.29	84.81	84.68	82.46	83.48	83.78	1.52
VH	86.07	84.54	86.12	85.73	85.70	85.82	85.15	86.37	86.18	86.59	85.83	0.60
WL	83.63	80.70	86.12	82.87	86.26	87.29	86.65	83.68	87.89	85.95	85.10	2.28

The remaining files

The remaining files were used to generate the specific plots in the review or to submit jobs on a computing cluster.

Contributors

This effort was done by the following individuals: Elisabetta Ghisu, Felipe Llinares-López, Philipp Nikolaus, Leslie O’Bray, and Bastian Rieck.