Randomly generated polytopes for testing mathematical programming algorithms
Randomly generated polytopes are used frequently to test and compare algorithms for a variety of mathematical programming problems. These polytopes are constructed by generating linear inequality constraints with coefficients drawn independently from a distribution such as the uniform or the normal. It is noted that this class of 'random' polytopes has a special property: the angles between the hyperplanes, though dependent on the specific distribution used, tend to be equal when the dimension of the space increases. Obviously this structure of 'random' polytopes may bias test results.
|Keywords||linear inequalities, random polytopes, testing and comparing algorithms|
|Persistent URL||dx.doi.org/10.1007/BF02592053, hdl.handle.net/1765/11647|
van Dam, W.B., Frenk, J.B.G., & Telgen, J.. (1983). Randomly generated polytopes for testing mathematical programming algorithms. Mathematical Programming, 26(2), 172–181. doi:10.1007/BF02592053