The first author is very grateful for the valuable discussion and help of Robert Irwin in converting and formatting the registration data prior to the scheduling process. We also would like to thank Andrew Gee and Martin Simmen for the useful comments and suggestions, and Carsten Peterson for the pointers and comments about his papers. Many thanks go to Karen Bedard for providing us with the data and answering so many questions we had about it, Meg Cortese for providing us with a set of building constraints for various departments, and Prof. Ben Ware, Vice President for Research and Computing at Syracuse University, for his support and encouragement.
Table 1: Size of the data set for each of the three semesters.
Table 2: The sparseness ratios of the problem for the data sets for each of
the three semesters. Lower values indicate a harder problem.
Table 3: Simulated annealing (SA) results: percentage of scheduled
classes, averaged over 10 runs of the same initial temperature and
other parameters for three terms. Expert system (ES) results:
percentage of scheduled classes is averaged over 10 runs. Mean-field
annealing (MFA) results: percentage of scheduled classes is also
averaged over 10 runs. No preprocessor was used with the
three methods.
Table 4: Percentage of scheduled classes, averaged over 10 runs of the
same initial temperature and other parameters, for three terms using
simulated anealing with an expert system as preprocessor.