View : 397 Download: 0

Full metadata record

DC Field Value Language
dc.contributor.advisor김선영-
dc.contributor.author김민준-
dc.creator김민준-
dc.date.accessioned2022-08-04T16:32:06Z-
dc.date.available2022-08-04T16:32:06Z-
dc.date.issued2022-
dc.identifier.otherOAK-000000191524-
dc.identifier.urihttps://dcollection.ewha.ac.kr/common/orgView/000000191524en_US
dc.identifier.urihttps://dspace.ewha.ac.kr/handle/2015.oak/261922-
dc.description.abstractQuadratic assignment problems (QAPs) are known as one of the most challenging problems in combinatorial optimization as their size frequently exceeds limits of most advanced methods and softwares developed. While a variety of methods has been proposed to solve the problems, one successful approach has been reduction techniques, more precisely, the facial and symmetry reduction techniques which is applied to semidefinite programming (SDP) relaxations of the QAPs. We discuss the effect of applying both facial and symmetry reduction together within an alternating direction method of multipliers (ADMM), a work by Oliveira et al. The problems that can be handled by their proposed work is limited in the sense that they require a specific form. We extend the types of problems that can be solved by the two reduction techniques. In particular, when the graph structure of cost matrices of QAPs forms Hamming graphs, we show that the ADMM outperforms other methods for SDP relaxations with the facial and symmetry reduction. The efficiency and stability of the ADMM to solve facially and symmetry reduced QAPs can be confirmed by the numerical experiment results on QAPs with cost matrices forming random Hamming graphs. ;이차 할당 문제(Quadratic assignment problems)는 문제의 사이즈가 해결 가능한 범주를 넘어서는 경우가 많기 때문에 조합 최적화(Combinatorial optimization)에서 해결하기 어려운 문제 중 하나로 알려져 있다. 문제를 해결하기 위해 연구된 다양한 방법들 중에서, 두 가지 차원 축소 방법-facial and symmetry reduction-은 성공적으로 이차 할당 문제의 준정부호 계획법(Semidefinite programmings)을 해결할 수 있다. Facial reduction과 symmetry reduction을 동시에 적용한 문제는 Oliveira et al에 의해 연구된 alternating direction method of multipliers (ADMM)로 효과적으로 해결할 수 있다. 특히 이차 할당 문제의 비용 행렬이 특정한 구조를 만족하는 경우에 위에 소개된 ADMM을 통해 해결할 수 있다. 본 논문에서는 두 가지 차원 축소 방법으로 해결할 수 있는 문제의 형태를 확장한다. 특히, 이차 할당 문제의 비용 행렬이 해밍 그래프(Hamming graphs)로 구성될 때, ADMM은 다른 준정부호 계획법 완화 문제를 해결하는 방법보다 뛰어난 결과를 제시한다. 이를 다양한 크기의 무작위 해밍 그래프를 비용 행렬로 가지는 이차 할당 문제에 적용하여, 효율성과 안정성의 향상을 확인하였다.-
dc.description.tableofcontents1. Introduction 1 2. Preliminaries 3 2.1. Notation and symbols 3 2.2. Semidefinite programmings and Quadratic assignment problems 3 2.2.1. Semidefinite programmings 3 2.2.2. Quadratic assignment problems 4 2.3. Hamming graphs 5 3. Dimension reduction and ADMM 6 3.1. Dimension reduction 6 3.1.1. Facial reduction 6 3.1.2. Facial reduction for SDPs 8 3.1.3. Symmetry reduction 9 3.1.4. Facial reduction for symmetry reduced SDPs 13 3.2. ADMM 15 3.2.1. Constructing subproblems 15 3.2.2. Solving subproblems 16 4. Facial reduction for symmetry reduced QAPs 19 4.1. Facial reduction for symmetry reduced QAPs 19 4.2. Solving QAPs with ADMM 22 5. Generalized QAPs with random Hamming graphs 26 6. Numerical results 28 6.1. Simple QAPs 28 6.2. Generalized QAPs 31 7. Conclusion 33 References 34-
dc.formatapplication/pdf-
dc.format.extent3079279 bytes-
dc.languageeng-
dc.publisher이화여자대학교 대학원-
dc.subject.ddc500-
dc.titleSolving Quadratic Assignment Problems based on Hamming Graphs with Reduction Techniques-
dc.typeMaster's Thesis-
dc.creator.othernameKim, Min Jun-
dc.format.pageiii, 37 p.-
dc.identifier.thesisdegreeMaster-
dc.identifier.major대학원 수학과-
dc.date.awarded2022. 8-
Appears in Collections:
일반대학원 > 수학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

BROWSE