On regularity of Max-CSPs and Min-CSPs
| Autor: | Aleksa Stankovic |
|---|---|
| Rok vydání: | 2022 |
| Předmět: |
FOS: Computer and information sciences
Computer Science - Computational Complexity TheoryofComputation_MATHEMATICALLOGICANDFORMALLANGUAGES TheoryofComputation_LOGICSANDMEANINGSOFPROGRAMS Signal Processing Computational Complexity (cs.CC) Computer Science::Artificial Intelligence Computer Science::Computational Complexity Computer Science::Data Structures and Algorithms Computer Science Applications Information Systems Theoretical Computer Science |
| Zdroj: | Information Processing Letters. 176:106244 |
| ISSN: | 0020-0190 |
| DOI: | 10.1016/j.ipl.2022.106244 |
| Popis: | We study approximability of regular constraint satisfaction problems, i.e., CSPs where each variable in an instance has the same number of occurrences. In particular, we show that for any CSP $\Lambda$, existence of an $\alpha$ approximation algorithm for unweighted regular Max-CSP $\Lambda$ implies existence of an $\alpha-o(1)$ approximation algorithm for weighted Max-CSP $\Lambda$ in which regularity of the instances is not imposed. We also give an analogous result for Min-CSPs, and therefore show that up to arbitrarily small error it is sufficient to conduct the study of approximability of CSPs only on regular unweighted instances. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect