| Tuning scenario | Training set of benchmark instances | Test set of benchmark instances | Instane/seed lists used for the repetitions of ParamILS (for the AAAI-07 article, we used the first 25) |
Citation | Short description |
| SW-GCPsat4j (all 20K SWGCP instances, 842 MB, .tar.gz) SAT4J params.txt file Note the forbidden parameter combinations: these combinations are disallowed since they can yield an incorrect SAT solver. Also note that special care needs to be taken care of the computational environment when tuning SAT4J to ensure the timing information is correct: based on Java, SAT4J might use more than one processor at a time, and it is thus important to ensure no other jobs are running on the machines used for tuning. |
Training set (names of 1000 instances) |
Test set (names of 1000 instances) |
Training instance/seed lists Test instance/seed list |
@inproceedings{GenHooProWal99, author = "I.~P. Gent and H.~H.~Hoos and P.~Prosser and T.~Walsh", title = "Morphing: Combining Structure and Randomness", booktitle = aaai99, pages = "654--660", year = "1999" } |
SAT-encoded graph colouring based on small world graphs. |
| SW-GCP-saps (all 20K SWGCP instances, same as above, 842 MB, .tar.gz) |
Training set (names of 113 instances) |
Test set (names of 100 instances) |
Training instance/seed lists Test instance/seed list |
@inproceedings{GenHooProWal99, author = "I.~P. Gent and H.~H.~Hoos and P.~Prosser and T.~Walsh", title = "Morphing: Combining Structure and Randomness", booktitle = aaai99, pages = "654--660", year = "1999" } |
SAT-encoded graph colouring based on small world graphs. |
| QWH-SAPS | Single instance, same for train and test | Training instance/seed lists Test instance/seed list |
@inproceedings{gomes97problem, lauthor = "Carla P. Gomes and Bart Selman", author = "C.~P. Gomes and B. Selman", title = "Problem Structure in the Presence of Perturbations", booktitle = aaai97, publisher = aaaimit, year = "1997", pages = {221--226}, editor = {B. Kuipers and B. Webber} } |
SAT-encoded quasigroup with holes instance | |
| GRID (all 100 grid instances, 2.5 MB, .tar.gz) GLS4MPE params.txt file |
Training set (names of 50 instances) |
Test set (names of 50 instances) |
Training instance/seed lists Test instance/seed list |
@MASTERSTHESIS{Hut04, lauthor = {Frank Hutter}, author = {F. Hutter}, title = {Stochastic Local Search for Solving the Most Probable Explanation Problem in {B}ayesian Networks}, school = {Department of Computer Science, Darmstadt University of Technology}, address = {Darmstadt, Germany}, year = {2004}, month = {September} } |
Grid-structured Bayesian networks with 25×25 binary variables, random conditional probability tables and 10 evidence variables |
Please send any questions, concerns or comments to Frank Hutter