Fundamental studies on methods for analyzing largescale combinatorial systems with submodular structures
Project/Area Number  08680450 
Research Category 
GrantinAid for Scientific Research (C)

Allocation Type  Singleyear Grants 
Section  一般 
Research Field 
社会システム工学

Research Institution  Osaka University(1997) University of Tsukuba(1996) 
Principal Investigator 
FUJISIGE Satoru Osaka Univ., DEPT.of CHEM.SCI.ENG., professor, 大学院・基礎工学研究科, 教授 (10092321)

Project Period (FY) 
1996 – 1997

Project Status 
Completed(Fiscal Year 1997)

Budget Amount *help 
¥2,600,000 (Direct Cost : ¥2,600,000)
Fiscal Year 1997 : ¥600,000 (Direct Cost : ¥600,000)
Fiscal Year 1996 : ¥2,000,000 (Direct Cost : ¥2,000,000)

Keywords  Submodular functions / Combinatorial optimization 
Research Abstract 
Submodular structures often play fundamental and important roles in a lot of problems related to discrete, combinatorial and algebraic systems such as graphs and networks through matroidal structures and flow boundary polyhedra. In the present research project we have investigated the fundamental properties of submodular structures that are useful in analyzing largescale discrete systems with submodular structures, and developed algorithms for solving problems for such discrete systems. The submodular flow problem is one of the most powerful mathematical models for optimization problems with submodular structures and its solution algorithms have been intensively investigated through the world. In the present research project we extended the push/relabel technique for network flows of A.V.Goldberg to the submodular flow problem and showed the effectiveness of the algorithm. We also investigated a combinatorial system, called bisubmodular system, that generalizes the combinatorial system
… More
with submodular structures. We obtained several useful results. First, we showed each domain of a bisubmodular function can be expressed as a collection of ideals of a signed poset. We proposed an efficient algorithm for finding a minimumweight ideal of a poset. We also developed an efficient algorithm for decomposing a bidirected graph into strongly connected components and showed the poset structure on the set of strongly connected components. Secondly, we revealed the fundamental properties of bisubmodular functions and associated polyhedra called bisubmodular polyhedra. In particular, we showed a minmax theorem with integrality for characterizing the mimimumnorm point, relative to l_1 norm, on a bisubmodular, polyhedron. This is an extremely powerful theorem with integrality that includes as special cases a lot of combinatorial minmax theorems such as intersection theorem for matroids, polymatroids and submodular systems. We expect that it will play an important role in dealing with various combinatorial optimization problems. We also revealed the relationship between bidirected network flows and bisubmodular systems. Less

Report
(3results)
Research Output
(25results)