by Ingo Althöfer, Franziska Berger, Stefan Schwarz
Preprint series: 02-04 , Reports on Optimization
Abstract: In many applications k-best algorithms produce only micro
mutations of the optimal solution instead of
true alternatives. A penalty method gives better chances to
find true alternatives.
For minimization problems with sum-type objective function
this approach generates alternatives whose
dissimilarity to the original optimum grows monotonically
with the penalty parameter.
Keywords: Multiple Choice System, k-best algorithm, micro mutation, true alternative, sum-type problem