摘要：Multidimensionalscalingaddressesthe problemhowmultidimensionaldatacanberepresented bypointsinalowdimensionalspace.Theproblem is reducedtoglobalminimizationofastressfunction whichmeasuresafitofdissimilaritybythedistances betweentherespectivepoints.Symmetriesindatamay exist.Performanceofglobaloptimizationmaybein- creasedreducingsearchspacesothatonlyoneofthe symmetric solutions should be found. Restriction of searchspaceisproposedanddemonstratedongeomet- ric datasetsformultidimensionalscaling.
