| 研究課題/領域番号 |
18F18117
|
|---|
| 研究機関 | 東京大学 |
|---|
研究代表者 |
陶山 明 東京大学, 大学院総合文化研究科, 教授 (90163063)
|
|---|
| 研究分担者 |
BARISH ROBERT 東京大学, 総合文化研究科, 外国人特別研究員
|
|---|
| 研究期間 (年度) |
2018-04-25 – 2020-03-31
|
| キーワード | Counting complexity / Graph theory / Combinatorics / Network analysis / Cycles / Paths / 2-Factors |
|---|
| 研究実績の概要 |
Towards a characterization of the complexity of cycle and path counting problems on graph classes arising in the context of, for example, biological network analysis and mathematical physics, we have shown (under standard complexity theoretic assumptions) that no combination of planarity, bipartiteness, vertex degree, and vertex connectivity constraints can conspire to allow for exact or approximate counting of Hamiltonian cycles, Hamiltonian paths, simple cycles, or simple paths in polynomial time. We also report significant progress on a program of using Courcelle's well-known metatheorem to automatically extend these intractability results to a large number of other natural graph classes. In the process of these efforts we have made a number of more general mathematical accomplishments, for example, in disproving a circa 2011 conjecture of Gordon et. al. that the Hamiltonian cycle decision problem on locally connected graphs of vertex degree at most 6 is efficiently solvable.
Concerning tractability results, for graph classes of bounded pathwidth we have established general methods for finding and proving generating functions, and where possible, analytic expressions, for counting Hamiltonian cycles, Hamiltonian paths, simple cycles, and simple paths. These methods succeed even in high-complexity cases, such as for infinite families of sliced recursive triangulations which arise in the context of mathematical physics models (e.g. in Lorentzian models). We have presented these findings at international mathematical physics conferences.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
As expected, we were able to characterize the hardness of counting and approximately counting Hamiltonian cycles and Hamiltonian paths on upwards of approximately 800 graph classes, including all 733 cases where there exists an NP-completeness result for the Hamiltonian cycle decision problem. For the majority of these cases, we have likewise been able to characterize the hardness of counting and approximately counting simple cycles and simple paths. Much more unexpectedly, however, was the fact that we were able to use Courcelle's well-known metatheorem to automate the process of establishing these hardness results for approximately 424 of aforementioned graph classes.
Another unexpected development was our ability to largely automate the process of finding and proving generating functions, and where possible, analytic expressions, for exactly counting Hamiltonian cycles, Hamiltonian paths, simple cycles, and simple paths on bounded pathwidth graphs of sufficiently high complexity to be of relevance to the mathematical physics community.
Additionally, we found a number of surprising connections between our work on cycle and path counting and the closely related problem of counting what are known as 2-factors of a graph (i.e. 2-regular covering subgraphs). These are related to what are known as fully-packed loop configurations in the mathematical physics community, and have close connections to six-vertex or ice-type models in statistical mechanics. We have already obtained a number of results for 2-factor counting problems on graph classes of interest to physicists.
|
| 今後の研究の推進方策 |
We will finalize and publish a number of manuscripts in preparation from 2018. This includes the proof, discussed in our summary of research achievements for FY2018, that no combination of planarity, bipartiteness, vertex degree, and vertex connectivity constraints can conspire to allow for exact or approximate counting of Hamiltonian cycles, Hamiltonian paths, simple cycles, or simple paths in polynomial time.
We will work towards completing our characterization of the complexity of exactly and approximately counting Hamiltonian cycles, Hamiltonian paths, simple cycles, and simple paths on all approximately 1246 classes of graphs in the ISGCI, or to show that these results would imply the solution to well-known open problems in the literature. On the basis of our accomplishments this year, we feel that this objective is within reach.
Additionally, as stated in the current status of the research section, there exist significant connections between our work on counting cycles and paths and the problem in mathematical physics of counting 2-factors on various types of graphs. Unfortunately, however, almost all complexity theoretic results for counting 2-factors are via reduction from the problem of counting Eulerian circuits, for which very little appears to be known. we will make an effort to build a bridge between our cycle and path counting results and 2-factor counting problems on a wide variety of graph classes, which if possible, would substantially improve the situation.
|
