GraphKernelBenchmark

Evaluation and comparison tool for different graph kernels

Installation

conda env create -f environment.yml
source activate graph_kernels

Setup

Navigate to /kernels/MLG/MLGkernel and edit path to Eigen in Makefile.options (even if it is included globally, it could be in usr/local/include or in usr/include)

make all

Navigate to /kernels/glocalwl and use

cmake ./
make

Run

Set -d parameter to folder names in datasets directory. In config.py you can find all currently supported kernels with their keys.

usage: evaluate.py [-h] -d DATA [DATA ...] -k KERNELS [KERNELS ...]

Starting benchmark

optional arguments:
  -h, --help            show this help message and exit
  -d DATA [DATA ...], --data DATA [DATA ...]
                        Benchmark datasets
  -k KERNELS [KERNELS ...], --kernels KERNELS [KERNELS ...]
                        Benchmark kernels