AL
Uploaded byAnna Lasota
PDF, PPTX265 views

MinFill_Presentation

This document discusses the minimum fill-in problem for sparse matrices. It begins with an introduction to fill-in that occurs during Gaussian elimination due to the introduction of new non-zero elements. It describes how the minimum fill-in problem is NP-hard and discusses various heuristics to minimize fill-in, including minimum degree ordering and nested dissection. The minimum degree algorithm works by repeatedly eliminating the vertex with minimum degree but does not always produce optimal orderings. The document provides examples to illustrate minimum degree and discusses enhancements like mass elimination to improve its performance.

Embed presentation

Download as PDF, PPTX
1 / 127
2 / 127
3 / 127
4 / 127
5 / 127
6 / 127
7 / 127
8 / 127
9 / 127
10 / 127
11 / 127
12 / 127
13 / 127
14 / 127
15 / 127
16 / 127
17 / 127
18 / 127
19 / 127
20 / 127
21 / 127
22 / 127
23 / 127
24 / 127
25 / 127
26 / 127
27 / 127
28 / 127
29 / 127
30 / 127
31 / 127
32 / 127
33 / 127
34 / 127
35 / 127
36 / 127
37 / 127
38 / 127
39 / 127
40 / 127
41 / 127
42 / 127
43 / 127
44 / 127
45 / 127
46 / 127
47 / 127
48 / 127
49 / 127
50 / 127
51 / 127
52 / 127
53 / 127
54 / 127
55 / 127
56 / 127
57 / 127
58 / 127
59 / 127
60 / 127
61 / 127
62 / 127
63 / 127
64 / 127
65 / 127
66 / 127
67 / 127
68 / 127
69 / 127
70 / 127
71 / 127
72 / 127
73 / 127
74 / 127
75 / 127
76 / 127
77 / 127
78 / 127
79 / 127
80 / 127
81 / 127
82 / 127
83 / 127
84 / 127
85 / 127
86 / 127
87 / 127
88 / 127
89 / 127
90 / 127
91 / 127
92 / 127
93 / 127
94 / 127
95 / 127
96 / 127
97 / 127
98 / 127
99 / 127
100 / 127
101 / 127
102 / 127
103 / 127
104 / 127
105 / 127
106 / 127
107 / 127
108 / 127
109 / 127
110 / 127
111 / 127
112 / 127
113 / 127
114 / 127
115 / 127
116 / 127
117 / 127
118 / 127
119 / 127
120 / 127
121 / 127
122 / 127
123 / 127
124 / 127
125 / 127
126 / 127
127 / 127
Minimum Fill-In Anna Lasota Berlin, December 2003
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Sparse Systems of Linear Equations Ax = b sparse, symmetric, positive definite matrices arise frequently in physical applications such as • finite element method • electrical network analysis • analysis of structural systems famous direct methods to solve a linear system Ax = b are • Gauss Elimination • Cholesky Factorization
Example Finite Element Method - Bath Tub
Solving Ax = b with Gauss   a11 . . . a1n ... ... ... an1 ann     x1 ... xn   =   b1 ... bn   eliminate first column a11x1 + · · · + a1nxn = b1 ⇐⇒ x1 = (−1 (a12x2 + · · · + a1nxn) + b1)/a11 substitute x1 in subsequent equations a21x1 + . . . + a2nxn = b2 ... ... ... ... an1x1 + . . . + annxn = bn
Introduction of Fill on a 3 × 3 Matrix   a11 a12 a13 a21 a22 a31 a33     x1 x2 x3   =   b1 b2 b3   eliminate first column a11x1 + a12x2 + a13x3 = b1 ⇐⇒ x1 = (−1 (a12x2 + a13x3) + b1)/a11 substitute x1 in subsequent equations a21x1 + a22x2 + = b2 a31x1 + + a33x3 = b3 =⇒   a22 a23 a32 a33     x2 x3   =   b2 b3   new non-zero elements a23 and a32 are called Fill-In.
Gauss Elimination Process and the Graph Theoretic View 1 2 3 4 × × × ×××× × × × 2 3 4 ××× × × × ×• • • • • • 3 4 ××× × × × × •• • • 4 ××× × × × × • ••
Disadvantages of Fill-In Fill-In increases the storage requirement Fill-In increases the arithmetic operations needed to solve the system Questions that arise: 1. Are there instances where no fill is introduced? 2. Is there a way to minimize the Fill-In introduced during elimination ?
Modelling the Elimination Process on a Graph Let the graph G = (V, E) correspond to A and let the vertex vi correspond to xi. Theorem: Elimination Graph [Parter 1961] Eliminating xi from the subsequent equations, the new graph Gvi of the remaining system is formed by: 1. eliminating the vertex vi that corresponds to xi from Gvi−1 2. pair-wise connecting the vertices of N(vi). v1 v2 v2 v3 v3 v3 v4v4 v4v4 G Gv1 Gv2 Gv3 →→ →
Modelling the Elimination Process on a Graph Proof: assume aii = 0, then xi = −1/aii k=i aikxk + yi/aii if aij = 0 no substitution need be made if aij = 0 and aik = 0 then • introduce xj to the kth equation and • introduce xk to the jth equation
EliminationGraph[Parter1961] × × × × × × × × × × ×× × × × × i ii i ii j j j j jj k k k k kk ............ ... ... ... ... =⇒
Solving Ax = b with Cholesky We can decompose A into A = LLT We know aij =< li, lj > constructing L • Algorithm at blackboard A = LLT = LU, where LT = U
Cholesky Factorization and Fill-In × × × × × × × ××× A = 1 2 3 4 × × × × ××× 1 2 3 4 × × × × ××× •→ 1 2 3 4 × × × × ××× • •→ 1 2 3 4 × × × ×× ×× • ••→ 1 2 3 4 × × × × ××× • •• U =
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Perfect Elimination Definition: [Rose 1972] A is a perfect elimination matrix if there exists a permutation matrix P such that ˜A = PAPT = LLT and aij = 0 ⇒ lij = 0, i < j. Definition: For a graph G = (V, E) with |V | = n an ordering of V is a bijection σ : {1, 2, . . . , n} → V . 1 1 22 3 3 4 4 v1v1 v1v1 v2v2 v2v2 v3v3 v3v3 v4 v4v4 v4 × × × × × × × ×× × × × × × × × × × × × × × ×× × × × ×× × × ××× • •• → LT = LT =
Chordal Graph (Hanjal und Suranyi [1958]) Definition: A Graph G = (V, E) is chordal if every cycle of length ≥ 4 has a chord. A chord is an edge connecting two nonconsecutive vertices of the cycle. G1 chordal G2 not chordal G3 not chordal The class of chordal graphs is the class of perfect elimination graphs.
Definition: The ordered graph Gσ is monotone transitive when ∀v ∈ V (v, u1) ∈ E and (v, u2) ∈ E =⇒ (u1, u2) ∈ E, with σ−1 (v) < σ−1 (u1) and σ−1 (v) < σ−1 (u2) Theorem: (Rose [1972]) For G = (V, E) the following statements are equivalent: A is a perfect elimination matrix. ∃ an ordering σ of V such that Gσ is monotone transitive σ is a perfect elimination order G is chordal
What happens if G is not chordal ? v1 v2 v3 v4 v5 v6 G
Elimination Graph v1 v2 v3 v4 v5 v6 G
Elimination Graph v1 v2 v3 v4 v5 v6 G
Elimination Graph v2 v3 v4 v5 v6 Gv1
Elimination Graph v2 v3 v4 v5 v6 Gv1
Elimination Graph v2 v3 v4 v5 v6 Gv1
Elimination Graph v2 v3 v4 v5 v6 Gv1
Elimination Graph v2 v3 v4 v5 v6 Gv1
Elimination Graph v2 v3 v4 v5 v6 Gv1
Elimination Graph v3 v4 v5 v6 Gv2
Elimination Graph v3 v4 v5 v6 Gv2
Elimination Graph v3 v4 v5 v6 Gv2
Elimination Graph v3 v4 v5 v6 Gv2
Elimination Graph v3 v4 v5 v6 Gv2
Elimination Graph v4 v5 v6 Gv3
Elimination Graph v4 v5 v6 Gv3
Elimination Graph v5 v6 Gv4
Elimination Graph v6 Gv5
The filled graph G+ v1 v2 v3 v4 v5 v6 G+ σ1 The implicit ordering σ1 of the vertices causes a fill of 4. G+ σ1 = (V, E ∪ F) is the filled graph, where |F| is the size of the fill.
There’s a better ordering v1 v1v2 v2v3 v3 v4 v4v5 v5v6 v6 G+ σ2 G+ σ1
Chordal Completion G+ σ is monotone transitive Eliminate vertices from G+ σ according to the same order σ used to obtain G+ σ from G G+ = (V, E ∪ F) is chordal. G+ is the chordal completion of G v1 v1v2 v2v3 v3 v4 v4v5 v5v6 v6 G+ σ2 G+ σ1
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Problem Formulation Minimum Fill-In Problem Instance: Graph G = (V, E) Task: Find a set of edges F of smallest size such that G = (V, E ∪ F) is chordal. also known as Minimum chordal graph completion Problem
Minimum Fill-In is NP-hard SAT ↓ 3 - SAT ↓ Simple Max Cut ↓ Simple Optimal Linear Arrangement ⇒ Optimal Linear Arrangement ↓ Minimum Fill-In
Minimum Fill-In vs. Minimal Fill-In G minimal chordal completion minimum chordal completion • minimal chordal completion (inclusion-minimal set of edges) can be computed in poly- nomial time • can be far from minimum
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Minimum Degree Algorithm repeatedly finds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm, runs in O(nm+ ), where m+ = |E ∪ F|
Minimum Degree Algorithm 1 2 3 4 5 6 G
1. Iteration - MD 1 2 3 4 5 6 G
2. Iteration - MD 2 3 4 5 6 G
3. Iteration - MD 2 4 5 6 G
4. Iteration - MD 2 5 6 G
5. Iteration - MD 5 6 G
6. Iteration - MD 6 G
Advantages - Disadvantages Minimum Degree Algorithm Advantages Disadvantages fast algorithm does not recognize chordal graphs does not in general produce a minimal ordering triangulation produced is arbitrarily greater than a minimum triangulation
Example Let p ≤ q Kq v a1 a2 . . . apa3
Example Let p ≤ q Kq v a1 a3 . . . apa2
Example Let p ≤ q Kq v a1 a3 . . . apa2
Example Let p ≤ q Kq v a1 a3 . . . apa2 deg(v) = p deg(ai) = q + 1 deg(kj) = q − 1 + p
Example Kq v a1 a3 . . . apa2
Example Kq v a1 a3 . . . apa2
Example Kq v a1 a3 apa2
Example Kq a1 a3 apa2
Example Kq+p a1 a3 apa2
A better way Kq v a1 a3 . . . apa2
A better way Kq v a1 a3 . . . apa2
A better way Kq v a1 a3 . . . apa2
A better way Kq v a1 a3 . . . apa2 Turn N(a1) into a clique adding q edges.
A better way Kq+1 v a3 . . . apa2
A better way Kq+1 v a3 . . . apa2
A better way Kq+1 v a3 . . . ap
A better way Kq+1 v a3 . . . ap
A better way Kq+1 v ap
A better way Kq+1 v ap
A better way Kq+1 v
Performance Guarantee dotsdots total fill of p(p − 1)/2, O(p2 ) Kq+p a1 a3 apa2 total fill of q Kq+1 v
Minimum Degree Algorithm repeatedly finds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm
Quotientgraph and Mass Elimination 1. Quotientgraph - Supernodes and Enodes, Reachable Set 2. Mass Elimination - u, v ∈ V are indistinguishable in G if NG(u)∪{u} = NG(v)∪{v}
Minimum Degree Algorithm repeatedly finds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm
Multiple Elimination Determine the minimum degree Choose the vertices that have the minimum degree Mark their neighbors as unknown degree
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1. Iteration Multiple Elimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
2. Iteration Multiple Elimination 2 3 5 6 7 8 9 10 11 12 14 15
2. Iteration Multiple Elimination 2 3 5 6 7 8 9 10 11 12 14 15
2. Iteration Multiple Elimination 2 3 5 6 7 8 9 10 11 12 14 15
2. Iteration Multiple Elimination 2 3 5 6 7 8 9 10 11 12 14 15
2. Iteration Multiple Elimination 2 3 5 6 7 8 9 10 11 12 14 15
2. Iteration Multiple Elimination 2 3 5 6 7 8 9 10 11 12 14 15
3. Iteration Multiple Elimination 3 5 6 7 10 11 12 14
4. Iteration Multiple Elimination 5 6 7 10 11 12
4. Iteration Multiple Elimination 6 7 10 11 12
4. Iteration Multiple Elimination 6 7 10 11 12
4. Iteration Multiple Elimination 6 7 10 11 12
4. Iteration Mass Elimination
Minimum Degree Algorithm repeatedly finds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm
External and Approximate Degree For a set of vertices X ⊂ V , N(X) = v∈X N(v) − X external degree of X is |N(X)|, that is the number of adjacent vertices that belong to other supernodes. NOT the number of adjacent supernodes approximate degree computes an upper bound on the degrees inexpensively instead of computing the exact degrees
Enhancements Runtime Algorithm Runtime MD O(nm+ ) MD QuotientGraph O(n2 m) MMD O(n2 m) AMD O(nm)
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Minimum Deficiency The deficiency of v is the number of edges needed to turn the neighborhood of a vertex v into a clique The deficiency corresponds to the number of fill edges introduced by eliminating v Try to minimize the overall amount of fill by, at each step, choosing to eliminate a vertex with minimum deficiency
Minimum Deficiency Algorithm v1 v2 v3 v4 v5 v6
Minimum Deficiency Algorithm v1 v2 v3 v4 v5 v6 1 3 1 1 3 1
Minimum Deficinecy Algorithm v2 v3 v4 v5 v6 2 1 0 2 1
Minimum Deficiency Algorithm v2 v3 v5 v6 1 1 1 1
Minimum Deficiency Algorithm v3 v5 v6 0 0 0
Minimum Deficiency Algorithm v5 v6 0 0
Minimum Deficiency Algorithm v6 0
Advantages - Disadvantages Minimum Deficiency Algorithm Advantages Disadvantages good quality ordering implementation too expensive recognizes chordal graphs not always a minimal triangulation 1 1 1 1 1 11 33
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Nested Dissection [George 1973] Nested Dissection examines the graph as a whole before ordering • orders the vertices of the graph backwards • begins by deciding which vertices should be eliminated last works by selecting a balanced separator • a set of vertices, that when removed from the graph, partitions it into connected components vertices in the separator are placed last in the elimination order recursively orders each of the connected components
Nested Dissection Algorithm
Nested Dissection Algorithm
Nested Dissection Algorithm
Nested Dissection Algorithm
Nested Dissection Algorithm
Performance Bound At most O(n log n) fill for graphs with small separators (O( √ n))[Lipton, Rose, Tarja 1979] • planar graphs • graphs with bounded genus • graphs with bounded degree Approximation algorithm [Agraval, Klein, Ravi 1990] with performance bound O( √ d log4 n) • for bounded degree graphs Theoretical result [Gilbert 1989]: for any graph there exisits a ND algorithm whose fill is within O(d log n) of minimum, where d is the maximum degree.
Advantages - Disadvantages Nested Dissection Advantages Disadvantages good ordering for limited practical use solving Ax = b in parallel finding a minimum balanced separator is NP-hard no fast algorithms for finding approximate separators
Questions left What properties should the vertex separator have? What algorithms should be used to find the vertex separator? When should the nested recursion be halted? How should the separator vertices be ordered? In Practice When connected component small enough use Minimum Degree Algorithm
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Hybrid Algorithms take advantage of of the best characteristics of Minimum Degree and Nested Dissection using a few levels of separators to control the fill introduced by minimum degree orders Hendrickson and Rothberg[1996] is the current champion of ordering algorithms - BEND (Bruce and Ed’s Nested Dissection) algorithm has neither known worst case fill nor work analysis
BEND - Combining Nested Dissection and Minimum Degree compress the graph Nested Dissection finds a few levels of separators until connected components have size at most n/32 vertices • using a multilevel algorithm to find vertex separators, allowing some imbalance invoke Minimum Degree on the subgraphs finally apply Minimum Degree to reorder all separator vertices
Difference between Nested Dissection and BEND Nested Dissection BEND select a balanced separator use multilevel to find separator allowing some imbalance recursively order each subgraph recursively order each subgraph until small enough until n/32 vertices in separator are reorder separator vertices placed last in any order using Minimum Degree
Overview Introduction • What is Fill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
Approximation Algorithm Approximation Ratio is O(8 · OPT) G = (V, E) has minimum fill size k identify in G a kernel set of vertices A ⊆ V , |A| ≤ 4k • using ideas of the partition algorithm • triangulate G by adding only edges between vertices of A no prior knowledge of k
Summary Introduction • Fill-In • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm

More Related Content

PPTX
Backtracking
bysubhradeep mitra
 
PPTX
Prim's algorithm
byPankaj Thakur
 
DOCX
B.tech ii unit-4 material vector differentiation
byRai University
 
DOC
Unit 4 jwfiles
byNv Thejaswini
 
PDF
Density theorems for Euclidean point configurations
byVjekoslavKovac1
 
PDF
Iterative Compression
byASPAK2014
 
PDF
Paths and Polynomials
byASPAK2014
 
PPTX
Prims & kruskal algorithms
byAyesha Tahir
 
Backtracking
bysubhradeep mitra
 
Prim's algorithm
byPankaj Thakur
 
B.tech ii unit-4 material vector differentiation
byRai University
 
Unit 4 jwfiles
byNv Thejaswini
 
Density theorems for Euclidean point configurations
byVjekoslavKovac1
 
Iterative Compression
byASPAK2014
 
Paths and Polynomials
byASPAK2014
 
Prims & kruskal algorithms
byAyesha Tahir
 

What's hot

PPT
01 knapsack using backtracking
bymandlapure
 
PDF
Maths04
bysansharmajs
 
PPTX
Minimum spanning tree algorithms by ibrahim_alfayoumi
byIbrahim Alfayoumi
 
PPT
ADA - Minimum Spanning Tree Prim Kruskal and Dijkstra
bySahil Kumar
 
PDF
2-rankings of Graphs
byAndrew Kallmeyer
 
PDF
Estimates for a class of non-standard bilinear multipliers
byVjekoslavKovac1
 
PDF
A sharp nonlinear Hausdorff-Young inequality for small potentials
byVjekoslavKovac1
 
PDF
On maximal and variational Fourier restriction
byVjekoslavKovac1
 
PDF
Some Examples of Scaling Sets
byVjekoslavKovac1
 
PPTX
Prims and kruskal algorithms
bySaga Valsalan
 
PDF
Treewidth and Applications
byASPAK2014
 
PPT
Muchtadi
byNiranjan Patidar
 
PDF
Topological sorting
byAmit Kumar Rathi
 
PPT
5.1 greedy
byKrish_ver2
 
PDF
Minimum spanning tree
byAmit Kumar Rathi
 
PDF
William hyatt-7th-edition-drill-problems-solution
bySalman Salman
 
PDF
Variants of the Christ-Kiselev lemma and an application to the maximal Fourie...
byVjekoslavKovac1
 
PDF
Tensorizing Neural Network
byRuochun Tzeng
 
PDF
Solving the energy problem of helium final report
byJamesMa54
 
PDF
New data structures and algorithms for \\post-processing large data sets and ...
byAlexander Litvinenko
 
01 knapsack using backtracking
bymandlapure
 
Maths04
bysansharmajs
 
Minimum spanning tree algorithms by ibrahim_alfayoumi
byIbrahim Alfayoumi
 
ADA - Minimum Spanning Tree Prim Kruskal and Dijkstra
bySahil Kumar
 
2-rankings of Graphs
byAndrew Kallmeyer
 
Estimates for a class of non-standard bilinear multipliers
byVjekoslavKovac1
 
A sharp nonlinear Hausdorff-Young inequality for small potentials
byVjekoslavKovac1
 
On maximal and variational Fourier restriction
byVjekoslavKovac1
 
Some Examples of Scaling Sets
byVjekoslavKovac1
 
Prims and kruskal algorithms
bySaga Valsalan
 
Treewidth and Applications
byASPAK2014
 
Muchtadi
byNiranjan Patidar
 
Topological sorting
byAmit Kumar Rathi
 
5.1 greedy
byKrish_ver2
 
Minimum spanning tree
byAmit Kumar Rathi
 
William hyatt-7th-edition-drill-problems-solution
bySalman Salman
 
Variants of the Christ-Kiselev lemma and an application to the maximal Fourie...
byVjekoslavKovac1
 
Tensorizing Neural Network
byRuochun Tzeng
 
Solving the energy problem of helium final report
byJamesMa54
 
New data structures and algorithms for \\post-processing large data sets and ...
byAlexander Litvinenko
 

Viewers also liked

PDF
1
bymichan mylo
 
PDF
Preparing to drive a heavy haulage truck
byNational Heavy Haulage
 
PDF
This is how to save when you tour in China
byThe China Travel Company
 
DOCX
Ciclo de vida de la empresa
byyasmila yauyo vargas
 
PPTX
Ambientes interativos e os impactos nos usuários (Aula 1)
byAndre de Abreu
 
PPTX
Sistemas electorales
byelgranlato09
 
PPTX
Participacion politica 2015
byelgranlato09
 
PPTX
Indisciplinaridade
byAndre de Abreu
 
PDF
Findasense_Credentials_ 2016_V2 1
byPriya Roopani
 
PPTX
Os Bastidores da Web Semântica
byAndre de Abreu
 
PDF
Top 10 qualities of a great truck driver - infographic
byNational Heavy Haulage
 
1
bymichan mylo
 
Preparing to drive a heavy haulage truck
byNational Heavy Haulage
 
This is how to save when you tour in China
byThe China Travel Company
 
Ciclo de vida de la empresa
byyasmila yauyo vargas
 
Ambientes interativos e os impactos nos usuários (Aula 1)
byAndre de Abreu
 
Sistemas electorales
byelgranlato09
 
Participacion politica 2015
byelgranlato09
 
Indisciplinaridade
byAndre de Abreu
 
Findasense_Credentials_ 2016_V2 1
byPriya Roopani
 
Os Bastidores da Web Semântica
byAndre de Abreu
 
Top 10 qualities of a great truck driver - infographic
byNational Heavy Haulage
 

Similar to MinFill_Presentation

PDF
Numerical Solution of Linear algebraic Equation
bypayalpriyadarshinisa1
 
PDF
Numerical Linear Algebra for Data and Link Analysis.
byLeonid Zhukov
 
PDF
Polylogarithmic approximation algorithm for weighted F-deletion problems
byAkankshaAgrawal55
 
PPTX
All Perfect Elimination Orderings & Minimal Vertex Seperators
bynazlitemu
 
PPT
4900514.ppt
byArunachalamSelva
 
PDF
Numerical Methods in Civil engineering for problem solving
byDrLVPRASADMeesaragan
 
PPTX
NP Complete Problems -- Internship
bySeshagiri Rao Kornepati
 
PDF
Graph in Data Structure
byProf Ansari
 
PDF
Unger
byUri Unger
 
PPTX
Floyd warshall algorithm and it's applications
byparth22110808
 
PPT
Lec-35Graph - Graph - Copy in Data Structure
byAnil Yadav
 
PPT
Graph
byLovepreet Kaur
 
PPT
shortestwarshashortestwarshashortestwarshall.ppt
byPravashSahoo5
 
PDF
My presentation all shortestpath
byCarlostheran
 
PPTX
(floyd's algm)
byJothi Lakshmi
 
PDF
Matrix Completion Presentation
byMichael Hankin
 
PDF
Kernel for Chordal Vertex Deletion
byAkankshaAgrawal55
 
PPTX
Implementation of graphs, adjaceny matrix
byVatsalSharma64
 
PPTX
141205 graphulo ingraphblas
bygraphulo
 
PPTX
141222 graphulo ingraphblas
byMIT
 
Numerical Solution of Linear algebraic Equation
bypayalpriyadarshinisa1
 
Numerical Linear Algebra for Data and Link Analysis.
byLeonid Zhukov
 
Polylogarithmic approximation algorithm for weighted F-deletion problems
byAkankshaAgrawal55
 
All Perfect Elimination Orderings & Minimal Vertex Seperators
bynazlitemu
 
4900514.ppt
byArunachalamSelva
 
Numerical Methods in Civil engineering for problem solving
byDrLVPRASADMeesaragan
 
NP Complete Problems -- Internship
bySeshagiri Rao Kornepati
 
Graph in Data Structure
byProf Ansari
 
Unger
byUri Unger
 
Floyd warshall algorithm and it's applications
byparth22110808
 
Lec-35Graph - Graph - Copy in Data Structure
byAnil Yadav
 
Graph
byLovepreet Kaur
 
shortestwarshashortestwarshashortestwarshall.ppt
byPravashSahoo5
 
My presentation all shortestpath
byCarlostheran
 
(floyd's algm)
byJothi Lakshmi
 
Matrix Completion Presentation
byMichael Hankin
 
Kernel for Chordal Vertex Deletion
byAkankshaAgrawal55
 
Implementation of graphs, adjaceny matrix
byVatsalSharma64
 
141205 graphulo ingraphblas
bygraphulo
 
141222 graphulo ingraphblas
byMIT
 

MinFill_Presentation

  • 1.
  • 2.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 3.
    Sparse Systems ofLinear Equations Ax = b sparse, symmetric, positive definite matrices arise frequently in physical applications such as • finite element method • electrical network analysis • analysis of structural systems famous direct methods to solve a linear system Ax = b are • Gauss Elimination • Cholesky Factorization
  • 4.
    Example Finite ElementMethod - Bath Tub
  • 5.
    Solving Ax =b with Gauss   a11 . . . a1n ... ... ... an1 ann     x1 ... xn   =   b1 ... bn   eliminate first column a11x1 + · · · + a1nxn = b1 ⇐⇒ x1 = (−1 (a12x2 + · · · + a1nxn) + b1)/a11 substitute x1 in subsequent equations a21x1 + . . . + a2nxn = b2 ... ... ... ... an1x1 + . . . + annxn = bn
  • 6.
    Introduction of Fillon a 3 × 3 Matrix   a11 a12 a13 a21 a22 a31 a33     x1 x2 x3   =   b1 b2 b3   eliminate first column a11x1 + a12x2 + a13x3 = b1 ⇐⇒ x1 = (−1 (a12x2 + a13x3) + b1)/a11 substitute x1 in subsequent equations a21x1 + a22x2 + = b2 a31x1 + + a33x3 = b3 =⇒   a22 a23 a32 a33     x2 x3   =   b2 b3   new non-zero elements a23 and a32 are called Fill-In.
  • 7.
    Gauss Elimination Processand the Graph Theoretic View 1 2 3 4 × × × ×××× × × × 2 3 4 ××× × × × ×• • • • • • 3 4 ××× × × × × •• • • 4 ××× × × × × • ••
  • 8.
    Disadvantages of Fill-In Fill-Inincreases the storage requirement Fill-In increases the arithmetic operations needed to solve the system Questions that arise: 1. Are there instances where no fill is introduced? 2. Is there a way to minimize the Fill-In introduced during elimination ?
  • 9.
    Modelling the EliminationProcess on a Graph Let the graph G = (V, E) correspond to A and let the vertex vi correspond to xi. Theorem: Elimination Graph [Parter 1961] Eliminating xi from the subsequent equations, the new graph Gvi of the remaining system is formed by: 1. eliminating the vertex vi that corresponds to xi from Gvi−1 2. pair-wise connecting the vertices of N(vi). v1 v2 v2 v3 v3 v3 v4v4 v4v4 G Gv1 Gv2 Gv3 →→ →
  • 10.
    Modelling the EliminationProcess on a Graph Proof: assume aii = 0, then xi = −1/aii k=i aikxk + yi/aii if aij = 0 no substitution need be made if aij = 0 and aik = 0 then • introduce xj to the kth equation and • introduce xk to the jth equation
  • 11.
  • 12.
    Solving Ax =b with Cholesky We can decompose A into A = LLT We know aij =< li, lj > constructing L • Algorithm at blackboard A = LLT = LU, where LT = U
  • 13.
    Cholesky Factorization andFill-In × × × × × × × ××× A = 1 2 3 4 × × × × ××× 1 2 3 4 × × × × ××× •→ 1 2 3 4 × × × × ××× • •→ 1 2 3 4 × × × ×× ×× • ••→ 1 2 3 4 × × × × ××× • •• U =
  • 14.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 15.
    Perfect Elimination Definition: [Rose1972] A is a perfect elimination matrix if there exists a permutation matrix P such that ˜A = PAPT = LLT and aij = 0 ⇒ lij = 0, i < j. Definition: For a graph G = (V, E) with |V | = n an ordering of V is a bijection σ : {1, 2, . . . , n} → V . 1 1 22 3 3 4 4 v1v1 v1v1 v2v2 v2v2 v3v3 v3v3 v4 v4v4 v4 × × × × × × × ×× × × × × × × × × × × × × × ×× × × × ×× × × ××× • •• → LT = LT =
  • 16.
    Chordal Graph (Hanjalund Suranyi [1958]) Definition: A Graph G = (V, E) is chordal if every cycle of length ≥ 4 has a chord. A chord is an edge connecting two nonconsecutive vertices of the cycle. G1 chordal G2 not chordal G3 not chordal The class of chordal graphs is the class of perfect elimination graphs.
  • 17.
    Definition: The orderedgraph Gσ is monotone transitive when ∀v ∈ V (v, u1) ∈ E and (v, u2) ∈ E =⇒ (u1, u2) ∈ E, with σ−1 (v) < σ−1 (u1) and σ−1 (v) < σ−1 (u2) Theorem: (Rose [1972]) For G = (V, E) the following statements are equivalent: A is a perfect elimination matrix. ∃ an ordering σ of V such that Gσ is monotone transitive σ is a perfect elimination order G is chordal
  • 18.
    What happens ifG is not chordal ? v1 v2 v3 v4 v5 v6 G
  • 19.
  • 20.
  • 21.
  • 22.
  • 23.
  • 24.
  • 25.
  • 26.
  • 27.
  • 28.
  • 29.
  • 30.
  • 31.
  • 32.
  • 33.
  • 34.
  • 35.
  • 36.
    The filled graphG+ v1 v2 v3 v4 v5 v6 G+ σ1 The implicit ordering σ1 of the vertices causes a fill of 4. G+ σ1 = (V, E ∪ F) is the filled graph, where |F| is the size of the fill.
  • 37.
    There’s a betterordering v1 v1v2 v2v3 v3 v4 v4v5 v5v6 v6 G+ σ2 G+ σ1
  • 38.
    Chordal Completion G+ σ ismonotone transitive Eliminate vertices from G+ σ according to the same order σ used to obtain G+ σ from G G+ = (V, E ∪ F) is chordal. G+ is the chordal completion of G v1 v1v2 v2v3 v3 v4 v4v5 v5v6 v6 G+ σ2 G+ σ1
  • 39.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 40.
    Problem Formulation Minimum Fill-InProblem Instance: Graph G = (V, E) Task: Find a set of edges F of smallest size such that G = (V, E ∪ F) is chordal. also known as Minimum chordal graph completion Problem
  • 41.
    Minimum Fill-In isNP-hard SAT ↓ 3 - SAT ↓ Simple Max Cut ↓ Simple Optimal Linear Arrangement ⇒ Optimal Linear Arrangement ↓ Minimum Fill-In
  • 42.
    Minimum Fill-In vs.Minimal Fill-In G minimal chordal completion minimum chordal completion • minimal chordal completion (inclusion-minimal set of edges) can be computed in poly- nomial time • can be far from minimum
  • 43.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 44.
    Minimum Degree Algorithm repeatedlyfinds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm, runs in O(nm+ ), where m+ = |E ∪ F|
  • 45.
  • 46.
    1. Iteration -MD 1 2 3 4 5 6 G
  • 47.
    2. Iteration -MD 2 3 4 5 6 G
  • 48.
    3. Iteration -MD 2 4 5 6 G
  • 49.
    4. Iteration -MD 2 5 6 G
  • 50.
  • 51.
  • 52.
    Advantages - DisadvantagesMinimum Degree Algorithm Advantages Disadvantages fast algorithm does not recognize chordal graphs does not in general produce a minimal ordering triangulation produced is arbitrarily greater than a minimum triangulation
  • 53.
    Example Let p ≤q Kq v a1 a2 . . . apa3
  • 54.
    Example Let p ≤q Kq v a1 a3 . . . apa2
  • 55.
    Example Let p ≤q Kq v a1 a3 . . . apa2
  • 56.
    Example Let p ≤q Kq v a1 a3 . . . apa2 deg(v) = p deg(ai) = q + 1 deg(kj) = q − 1 + p
  • 57.
  • 58.
  • 59.
  • 60.
  • 61.
  • 62.
    A better way Kq v a1a3 . . . apa2
  • 63.
    A better way Kq v a1a3 . . . apa2
  • 64.
    A better way Kq v a1a3 . . . apa2
  • 65.
    A better way Kq v a1a3 . . . apa2 Turn N(a1) into a clique adding q edges.
  • 66.
  • 67.
  • 68.
  • 69.
  • 70.
  • 71.
  • 72.
  • 73.
    Performance Guarantee dotsdots total fillof p(p − 1)/2, O(p2 ) Kq+p a1 a3 apa2 total fill of q Kq+1 v
  • 74.
    Minimum Degree Algorithm repeatedlyfinds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm
  • 75.
    Quotientgraph and MassElimination 1. Quotientgraph - Supernodes and Enodes, Reachable Set 2. Mass Elimination - u, v ∈ V are indistinguishable in G if NG(u)∪{u} = NG(v)∪{v}
  • 76.
    Minimum Degree Algorithm repeatedlyfinds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm
  • 77.
    Multiple Elimination Determine theminimum degree Choose the vertices that have the minimum degree Mark their neighbors as unknown degree
  • 78.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 79.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 80.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 81.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 82.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 83.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 84.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 85.
    1. Iteration MultipleElimination 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
  • 86.
    2. Iteration MultipleElimination 2 3 5 6 7 8 9 10 11 12 14 15
  • 87.
    2. Iteration MultipleElimination 2 3 5 6 7 8 9 10 11 12 14 15
  • 88.
    2. Iteration MultipleElimination 2 3 5 6 7 8 9 10 11 12 14 15
  • 89.
    2. Iteration MultipleElimination 2 3 5 6 7 8 9 10 11 12 14 15
  • 90.
    2. Iteration MultipleElimination 2 3 5 6 7 8 9 10 11 12 14 15
  • 91.
    2. Iteration MultipleElimination 2 3 5 6 7 8 9 10 11 12 14 15
  • 92.
    3. Iteration MultipleElimination 3 5 6 7 10 11 12 14
  • 93.
    4. Iteration MultipleElimination 5 6 7 10 11 12
  • 94.
    4. Iteration MultipleElimination 6 7 10 11 12
  • 95.
    4. Iteration MultipleElimination 6 7 10 11 12
  • 96.
    4. Iteration MultipleElimination 6 7 10 11 12
  • 97.
    4. Iteration MassElimination
  • 98.
    Minimum Degree Algorithm repeatedlyfinds a vertex of minimum degree and eliminates it Enhancements • Mass Elimination using Supernodes [George, Liu 1980] • Multiple Elimination [Liu 1985] • External Degree [Liu 1985] • Approximate Degree [Amestoy, Davis, Duff 1996] fast algorithm
  • 99.
    External and ApproximateDegree For a set of vertices X ⊂ V , N(X) = v∈X N(v) − X external degree of X is |N(X)|, that is the number of adjacent vertices that belong to other supernodes. NOT the number of adjacent supernodes approximate degree computes an upper bound on the degrees inexpensively instead of computing the exact degrees
  • 100.
    Enhancements Runtime Algorithm Runtime MDO(nm+ ) MD QuotientGraph O(n2 m) MMD O(n2 m) AMD O(nm)
  • 101.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 102.
    Minimum Deficiency The deficiencyof v is the number of edges needed to turn the neighborhood of a vertex v into a clique The deficiency corresponds to the number of fill edges introduced by eliminating v Try to minimize the overall amount of fill by, at each step, choosing to eliminate a vertex with minimum deficiency
  • 103.
  • 104.
    Minimum Deficiency Algorithm v1v2 v3 v4 v5 v6 1 3 1 1 3 1
  • 105.
    Minimum Deficinecy Algorithm v2v3 v4 v5 v6 2 1 0 2 1
  • 106.
  • 107.
  • 108.
  • 109.
  • 110.
    Advantages - DisadvantagesMinimum Deficiency Algorithm Advantages Disadvantages good quality ordering implementation too expensive recognizes chordal graphs not always a minimal triangulation 1 1 1 1 1 11 33
  • 111.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 112.
    Nested Dissection [George1973] Nested Dissection examines the graph as a whole before ordering • orders the vertices of the graph backwards • begins by deciding which vertices should be eliminated last works by selecting a balanced separator • a set of vertices, that when removed from the graph, partitions it into connected components vertices in the separator are placed last in the elimination order recursively orders each of the connected components
  • 113.
  • 114.
  • 115.
  • 116.
  • 117.
  • 118.
    Performance Bound At mostO(n log n) fill for graphs with small separators (O( √ n))[Lipton, Rose, Tarja 1979] • planar graphs • graphs with bounded genus • graphs with bounded degree Approximation algorithm [Agraval, Klein, Ravi 1990] with performance bound O( √ d log4 n) • for bounded degree graphs Theoretical result [Gilbert 1989]: for any graph there exisits a ND algorithm whose fill is within O(d log n) of minimum, where d is the maximum degree.
  • 119.
    Advantages - DisadvantagesNested Dissection Advantages Disadvantages good ordering for limited practical use solving Ax = b in parallel finding a minimum balanced separator is NP-hard no fast algorithms for finding approximate separators
  • 120.
    Questions left What propertiesshould the vertex separator have? What algorithms should be used to find the vertex separator? When should the nested recursion be halted? How should the separator vertices be ordered? In Practice When connected component small enough use Minimum Degree Algorithm
  • 121.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 122.
    Hybrid Algorithms take advantageof of the best characteristics of Minimum Degree and Nested Dissection using a few levels of separators to control the fill introduced by minimum degree orders Hendrickson and Rothberg[1996] is the current champion of ordering algorithms - BEND (Bruce and Ed’s Nested Dissection) algorithm has neither known worst case fill nor work analysis
  • 123.
    BEND - CombiningNested Dissection and Minimum Degree compress the graph Nested Dissection finds a few levels of separators until connected components have size at most n/32 vertices • using a multilevel algorithm to find vertex separators, allowing some imbalance invoke Minimum Degree on the subgraphs finally apply Minimum Degree to reorder all separator vertices
  • 124.
    Difference between NestedDissection and BEND Nested Dissection BEND select a balanced separator use multilevel to find separator allowing some imbalance recursively order each subgraph recursively order each subgraph until small enough until n/32 vertices in separator are reorder separator vertices placed last in any order using Minimum Degree
  • 125.
    Overview Introduction • What isFill-In ? • Perfect Elimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm
  • 126.
    Approximation Algorithm Approximation Ratiois O(8 · OPT) G = (V, E) has minimum fill size k identify in G a kernel set of vertices A ⊆ V , |A| ≤ 4k • using ideas of the partition algorithm • triangulate G by adding only edges between vertices of A no prior knowledge of k
  • 127.
    Summary Introduction • Fill-In • PerfectElimination and Chordal Graphs • The Minimum Fill-In Problem Heuristics • Minimum Degree • Minimum Deficiency • Nested Dissection • Hybrid Algorithm Approximation Algorithm