Abstract
Performing Gaussian elimination to a sparse matrix may turn some zeroes into nonzero values, so called fill-ins, which we want to minimize to keep the matrix sparse. Let n denote the rows of the matrix and k the number of fill-ins. For the minimum fill-in problem, we exclude the existence of polynomial time approximation schemes, assuming P6=NP, and the existence of 2O(n1)-time approximation schemes for any positive , assuming the Exponential Time Hypothesis. Also implied is a 2O(k1=2) nO(1) parameterized lower bound. Behind these results is a new reduction from vertex cover, which might be of its own interest: All previous reductions for similar problems are from some kind of graph layout problems.
Original language | English |
---|---|
Title of host publication | 28th Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2017 |
Publisher | Association for Computing Machinery |
Pages | 875-880 |
Number of pages | 6 |
ISBN (Electronic) | 9781611974782 |
Publication status | Published - 1 Jan 2017 |
Event | 28th Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2017 - Hotel Porta Fira, Barcelona, Spain Duration: 16 Jan 2017 → 19 Jan 2017 |
Conference
Conference | 28th Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2017 |
---|---|
Country/Territory | Spain |
City | Barcelona |
Period | 16/01/17 → 19/01/17 |
ASJC Scopus subject areas
- Software
- General Mathematics