Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4640166 | Journal of Computational and Applied Mathematics | 2011 | 12 Pages |
Abstract
We introduce a new kind of kernel function, which yields efficient large-update primal-dual interior-point methods. We conclude that in some situations its iteration bounds are O(m3m+12mnm+12mlognϵ), which are at least as good as the best known bounds so far, O(nlognlognϵ), for large-update primal-dual interior-point methods. The result decreases the gap between the practical behavior of the large-update algorithms and their theoretical performance results, which is an open problem. Numerical results show that the algorithms are feasible.
Keywords
Related Topics
Physical Sciences and Engineering
Mathematics
Applied Mathematics
Authors
Liying Liu, Shaoyong Li,