liu.seSearch for publications in DiVA
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
A Multivariate Complexity Analysis of Qualitative Reasoning Problems
Linköping University, Department of Computer and Information Science, Artificial Intelligence and Integrated Computer Systems. Linköping University, Faculty of Science & Engineering.
Linköping University, Department of Computer and Information Science, Artificial Intelligence and Integrated Computer Systems. Linköping University, Faculty of Science & Engineering.
2022 (English)In: Proceedings of the 31st International Joint Conference on Artificial Intelligence (IJCAI-22) / [ed] Luc De Raedt, International Joint Conferences on Artificial Intelligence , 2022, p. 1804-1810Conference paper, Published paper (Refereed)
Abstract [en]

Qualitative reasoning is an important subfeld ofartifcial intelligence where one describes relationships with qualitative, rather than numerical, relations. Many such reasoning tasks, e.g., ALLEN’SINTERVAL ALGEBRA, can be solved in 2O(n·log n) time, but single-exponential running times 2O(n) are currently far out of reach. In this paper we consider single-exponential algorithms via a multivariate analysis consisting of a fine-grained parameter n (e.g., the number of variables) and a coarsegrained parameter k expected to be relatively small.We introduce the classes FPE and XE of problems solvable in f(k) · 2O(n), respectively f(k)n,time, and prove several fundamental properties ofthese classes. We proceed by studying temporal reasoning problems and (1) show that the PARTIALLYORDERED TIME problem of effective width k is solvable in 16kn time and is thus included in XE, and (2) that the network consistency problem for ALLEN’S INTERVAL ALGEBRA with no intervaloverlapping with more than k others is solvable in (2nk)2k· 2n time and is included in FPE. Our multivariate approach is in no way limited to these tospecifc problems and may be a generally useful approach for obtaining single-exponential algorithms. 

Place, publisher, year, edition, pages
International Joint Conferences on Artificial Intelligence , 2022. p. 1804-1810
Series
Proceedings of the International Joint Conference on Artificial Intelligence, ISSN 1045-0823
Keywords [en]
Constraint Satisfaction and Optimization: Constraint Satisfaction
National Category
Computer Sciences Computational Mathematics
Identifiers
URN: urn:nbn:se:liu:diva-198450DOI: 10.24963/ijcai.2022/251ISI: 001202342301129Scopus ID: 2-s2.0-85137850976ISBN: 9781956792003 (electronic)OAI: oai:DiVA.org:liu-198450DiVA, id: diva2:1804538
Conference
International Joint Conference on Artificial Intelligence, Messe Wien, Vienna, Austria, July 23-29, 2022
Note

Funding Agencies|National Graduate School in Computer Science (CUGS), Sweden; Swedish Research Council (VR) [2019-03690]; Swedish Research Council [2019-03690] Funding Source: Swedish Research Council

Available from: 2023-10-13 Created: 2023-10-13 Last updated: 2025-10-13Bibliographically approved
In thesis
1. Infinite-Domain CSPs and QBF: Fine-Grained and Parameterized Complexity
Open this publication in new window or tab >>Infinite-Domain CSPs and QBF: Fine-Grained and Parameterized Complexity
2025 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

While we today have quite powerful tools for solving problems that are NP-hard, or even harder ones, it is typically easy to give conditions where they exhibit impractical slow performance. When designing new, better, algorithms for these cases, understanding theoretical limits becomes crucial to avoid investing time in approaches that are ultimately dead ends. Modern conjectures, such as the exponential time hypothesis (ETH), enable us to establish effective theoretical lower bounds for many problems. These lower bounds often align closely with our best-known upper bounds, especially in problems over finite domains. However, this alignment tends to break down in cases involving infinite domains, or input-dependent domains, and for problems beyond NP. While we for some easier and harder infinite-domain problems have matching upper and lower bounds, there exists a wide range of problems where a significant knowledge gap remains. We specifically examine Allen’s interval algebra (Allen) and partially ordered time (POT), where the best known lower bounds are 2o(n). Both these problems can be formulated as infinite-domain constraint satisfaction problems (CSP) and exhibit this gap between upper and lower bounds. While these problems are solvable in 2O(n2) time by exhaustive search, we improve upon this and ultimately reach the first o(n)n algorithm for Allen. This result is the usage of dynamic programming, with a particular emphasis on tracking unsolved subproblems, rather than the more traditional approach of building upon already-solved subproblems.

While a significant improvement over exhaustive search, to get closer to single-exponential running times of 2O(n2), we shift our focus to (multivariate) parameterized complexity. We begin by introducing two new single-exponential complexity classes: fixed parameter single-exponential (FPE) and slicewise single-exponential (XE), analogous to the well-known classes of fixed-parameter tractable (FPT) and slicewise polynomial (XP), respectively. We then apply these concepts to Allen and POT, showing both FPE and XE results.

In the latter part of the thesis we shift focus to a problem where further unconditional improvements are unlikely under the strong ETH: evaluating quantified Boolean formulas (QBF). Although this problem is the PSpace-complete analogue of the Boolean Satisfiability problem (SAT), it is comparatively understudied, and few positive algorithmic results are known. Focusing on how simplifying away a small set of variables (a backdoor) results in a tractable formula, we start by showing how removing all existential variables yields new FPT results. Building upon this, we then show multiple other backdoor results for classical tractable classes like 2-CNF, AFF and HORN, including both new hardness results and new FPT algorithms. 

Place, publisher, year, edition, pages
Linköping: Linköping University Electronic Press, 2025. p. 27
Series
Linköping Studies in Science and Technology. Dissertations, ISSN 0345-7524 ; 2471
National Category
Computer Sciences
Identifiers
urn:nbn:se:liu:diva-217673 (URN)10.3384/9789181182149 (DOI)9789181182132 (ISBN)9789181182149 (ISBN)
Public defence
2025-10-20, Ada Lovelace, B Building, Campus Valla, Linköping, 13:15 (English)
Opponent
Supervisors
Note

Funding agency: The National Graduate School of Computer Science in Sweden (CUGS)

Available from: 2025-09-12 Created: 2025-09-12 Last updated: 2025-10-27Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopusPaper

Authority records

Eriksson, LeifLagerkvist, Victor

Search in DiVA

By author/editor
Eriksson, LeifLagerkvist, Victor
By organisation
Artificial Intelligence and Integrated Computer SystemsFaculty of Science & Engineering
Computer SciencesComputational Mathematics

Search outside of DiVA

GoogleGoogle Scholar

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 240 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf