Reverse Search for Parametric Linear Programming
This paper introduces a new enumeration technique for (multi)parametric linear programs (pLPs) based on the reverse-search paradigm. We prove that the proposed algorithm has a computational complexity that is linear in the size of the output (number of so-called critical regions) and a constant space complexity. This is an improvement over the quadratic and linear computational and space complexities of current approaches. Current implementations of the proposed approach become faster than existing methods for large problems. Extensions of this method are proposed that make the computational requirements lower than those of existing approaches in all cases, while allowing for efficient parallelisation and bounded memory usage
04177566.pdf
Publisher's version
restricted
226.85 KB
Adobe PDF
015d7744224890132c3b55fc038cfdef
publication_2473.pdf
Preprint
openaccess
176.3 KB
Adobe PDF
886390ae9d9a83acbb601b554104093f