Extending Instructional Support in an Algorithm Benchmarking System

OptimEx is an interactive system aimed at empirically comparing the outcomes of different optimization algorithms constructed to solve the same problem. In this paper, several extensions intended to improve the educational use of OptimEx are presented. Firstly, experiment setting has been enhanced, by means of both the automatic load of test cases provided by the instructor and the automatic load of the value ranges of test cases to be randomly generated. The presentation of results also was improved, by means of more refined statistics, the identification of test cases with the largest deviations from optimal results, or the possibility to present deviations as absolute or relative differences. Last but not least, the system allows measuring and comparing run times of the algorithms, using the same scheme of random generation of test cases.