Supported problems:

• max clique (optimization & decision),
• subgraph isomorphism,
• maximum common induced subgraph and its encoding as max clique for labelled and unlabelled graphs,
• graph edit distance (GED):
  • a constraint programming model,
  • 2 encodings as minimum weight clique:
    • with weights on vertices,
    • with weights on both vertices and edges.

How to run a non-GED experiment

1. Set the relevant parameters at the top of Makefile.
2. Make sure the MiniZinc models listed as prerequisites of some of the targets are set correctly.
3. Run make target for your selected target. The results will be gradually written to either clique.csv or subgraph.csv or both, depending on the target.
4. Modify results/plotting.R as needed to extract meaningful information from the data.

How to run a GED experiment

1. Set the relevant parameters at the top of Makefile. DATABASE can be one of: CMU, GREC, Mutagenicity, Protein.
2. Run make clean if you're changing the INFO_FILE but not DATABASE, otherwise Step 3 will run experiments for both old and new data. Warning: it will remove the CSV files as well.
3. Run make convert-model, where model is one of: cp, vertex-edge-weights, vertex-weights, mwc. The first three are MiniZinc models, the last one is a C program. This converts the pairs of graphs mentioned in the INFO_FILE into either DZN or DIMACS format for the appropriate model. Note that if a generated file already exists, it will NOT be overwritten.
4. Run make model (for the same value of model) and results will be gradually written to model.csv. Use -j and -l flags to run multiple different problem instances in parallel.
5. Modify and use results/plotting2.R as needed to extract meaningful information from the data.

How to test the whole system

Some minimalistic testing can be run using make test. Before running the command, run make clean to remove data files so the algorithms don't get run on them, and edit the parameters at the top of the Makefile so that:

• parameters for non-GED problems are not too big, so the tests don't take too long;
• DATABASE is set to GREC;
• INFO_FILE is set to the GREC5 low level info file;
• INT_VERSION is disabled;
• SET_INCUMBENT is disabled. This will check if the necessary files are created and if they contain an appropriate number of commas in each line (where appropriate). CP model is currently not tested as it doesn't properly support floating-point numbers.

Additional comments

• If you run a make model command several times, the results file will be overwritten.
• Sometimes the graph generator can output data files that are clearly unsatisfiable. That may result in one of the following:
  • WARNING: model inconsistency detected,
  • Error: syntax error, unexpected |]. The latter case produces a half-empty line in the results file because the model cannot be run on a completely empty adjacency matrix. These errors can be safely ignored.