|
|
| (One intermediate revision by one other user not shown) |
| Line 1: |
Line 1: |
| In [[numerical analysis]] the '''minimum degree algorithm''' is an [[algorithm]] used to permute the rows and columns of a [[symmetric matrix|symmetric]] [[sparse matrix]] before applying the [[Cholesky decomposition]], to reduce the number of non-zeros in the Cholesky factor.
| | I'm Luther but I by no means really favored that name. Alabama has usually been his home. Bottle tops collecting is the only pastime his spouse doesn't approve of. Managing individuals is what I do in my working day occupation.<br><br>Here is my blog post :: auto warranty; [http://www.9Shapes.eu/index.php?mod=users&action=view&id=22096 mouse click the up coming web site], |
| This results in reduced storage requirements and means that the Cholesky factor, or sometimes an incomplete Cholesky factor used as a preconditioner (for example in the preconditioned [[conjugate gradient]] algorithm) can be applied with fewer arithmetic operations.
| |
| | |
| Minimum degree algorithms are often used in the [[finite element method]] where the reordering of nodes can be carried out depending only on the topology of the mesh, rather than the coefficients in the partial differential equation, resulting in efficiency savings when the same mesh is used for a variety of coefficient values.
| |
| | |
| Given a linear system
| |
| :<math> \mathbf{A}\mathbf{x} = \mathbf{b}</math>
| |
| where '''A''' is an <math>n \times n</math> real symmetric sparse square matrix the Cholesky factor '''L''' will typically suffer 'fill in', that is have more non-zeros than the upper triangle of '''A'''. We seek a [[permutation matrix]] '''P''', so that the matrix
| |
| <math>\mathbf{P}^T\mathbf{A}\mathbf{P}</math>, which is also symmetric, has the least possible fill in its Cholesky factor. We solve the reordered system
| |
| :<math> \left(\mathbf{P}^T\mathbf{A}\mathbf{P}\right) \left(\mathbf{P}^T\mathbf{x}\right) = \mathbf{P}^T\mathbf{b}.</math>
| |
| The problem of finding the best ordering is an [[NP-complete]] problem and is thus intractable, so heuristic methods are used instead. The minimum degree algorithm is derived from a method first proposed by [[Harry Markowitz|Markowitz]] in 1959 for non-symmetric [[linear programming]] problems, which is loosely described as follows. At each step in [[Gaussian elimination]] row and column permutations are performed so as to minimize the number of off diagonal non-zeros in the pivot row and column. A symmetric version
| |
| of Markowitz method was described by Tinney and Walker in 1967 and Rose later derived a graph theoretic version of the algorithm where the factorization is only simulated, and this was named the minimum degree algorithm. The graph referred to is the graph with ''n'' vertices, with vertices ''i'' and ''j'' connected by an edge when <math>a_{ij} \ne 0</math>, and the ''degree'' is the degree of the vertices. A crucial aspect of such algorithms is a tie breaking strategy when there is a choice of renumbering resulting in the same degree.
| |
| | |
| A version of the minimum degree algorithm was implemented in the [[MATLAB]] function '''symmmd''' (where MMD stands for multiple minimum degree), but has now been superseded by a symmetric approximate multiple minimum degree function '''symamd''', which is faster. This is confirmed by theoretical analysis, which shows that for graphs on ''n'' vertices and ''m'' edges, MMD has a tight [[Big O notation|upper bound]] of <math>O(n^2m)</math> on its run time, whereas for AMD a tight bound of <math>O(nm)</math> holds.
| |
| | |
| ==References==
| |
| *{{cite journal |authorlink=Harry Markowitz |first=H. M. |last=Markowitz |title=The elimination form of the inverse and its application to linear programming |journal=[[Management Science: A Journal of the Institute for Operations Research and the Management Sciences|Management Science]] |volume=3 |issue=3 |pages=255–269 |year=1957 |jstor=2627454 }}
| |
| *{{cite journal |first=Alan |last=George |first2=Joseph |last2=Liu |title=The evolution of the Minimum Degree Ordering Algorithm |journal=[[SIAM Review]] |volume=31 |issue=1 |pages=1–19 |year=1989 |jstor=2030845 }}
| |
| *{{cite journal |first=W. F. |last=Tinney |first2=J. W. |last2=Walker |title=Direct solution of sparse network equations by optimally ordered triangular factorization |journal=[[Proceedings of the IEEE|Proc. IEEE]] |volume=55 |issue=11 |pages=1801–1809 |year=1967 |doi=10.1109/PROC.1967.6011 }}
| |
| *{{cite book |first=D. J. |last=Rose |chapter=A graph-theoretic study of the numerical solution of sparse positive definite systems of linear equations |title=Graph Theory and Computing |editor-first=R. C. |editor-last=Read |location=New York |publisher=Academic Press |year=1972 |pages=183–217 |isbn=0-12-583850-6 }}
| |
| *{{citation|first=P.|last=Heggernes |first2=S. C. |last2=Eisenstat |first3=G. |last3= Kumfert| first4=A. |last4=Pothen |date=December 2001 |title=The Computational Complexity of the Minimum Degree Algorithm |type= Technical report |url=http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=ADA398632 |publisher=Institute for Computer Applications in Science and Engineering}}
| |
| | |
| {{Numerical linear algebra}}
| |
| | |
| {{DEFAULTSORT:Minimum Degree Algorithm}}
| |
| [[Category:Numerical linear algebra]]
| |
| [[Category:Matrix theory]]
| |
I'm Luther but I by no means really favored that name. Alabama has usually been his home. Bottle tops collecting is the only pastime his spouse doesn't approve of. Managing individuals is what I do in my working day occupation.
Here is my blog post :: auto warranty; mouse click the up coming web site,