The tiling by the minimal separators of a junction tree and applications to graphical models Durham, 2008

Size: px
Start display at page:

Download "The tiling by the minimal separators of a junction tree and applications to graphical models Durham, 2008"

Transcription

1 The tiling by the minimal separators of a junction tree and applications to graphical models Durham, 2008 G. Letac, Université Paul Sabatier, Toulouse. Joint work with H. Massam 1

2 Motivation : the Wishart distributions on decomposable graphs. We denote by S n the space of symmetric real matrices of order n and by P n S n the cone of positive definite matrices. Let G = (V, E) be a decomposable graph with V = {1,..., n}. The subspace ZS G S n is the space of symmetric matrices (s ij ) with zeros prescribed by G, that means s ij = 0 when {i, j} / E. We denote P G = ZS G P n. A space isomorphic to ZS G is the space IS G of symmetric incomplete matrices which are actually real functions on the union of the set V and the set E of edges. 2

3 We denote by π the natural projection of S n on IS G and denote Q G = π(pg 1 ). Three equivalent properties 1. Q G = π(pg 1 ) (definition) 2. Q G is the open convex cone which is the dual of the cone P G. 3. the restriction x C of x Q G to any clique C is positive definite (Hélène s definition). 3

4 Example : If G = the cone P G is the set of positive definite matrices of the form y 1 y 12 0 y 12 y 2 y 23 0 y 32 y 3 The cone Q G is the set of incomplete matrices of the form x 1 x 12 x 12 x 2 x 23 x 32 x 3 such that the two submatrices associated to the two cliques [ ] [ ] x1 x 12 x2 x, 23 x 12 x 2 x 32 x 3 are positive definite. 4

5 The bijection between P G and Q G. Let G decomposable, let C and S be the families of cliques and minimal separators. If x Q G define the Lauritzen function : y = ψ(x) = [x 1 C ] 0 C C S S ν(s)[x 1 S ] 0 where [a] 0 means extension by zeros of a principal submatrix a of S n and where ν(s) is a certain positive integer called multiplicity of S. Theorem 1. The map x y = ψ(x) is a diffeomorphism from Q G onto P G. Its inverse y x from P G onto Q G is x = π(y 1 ). 5

6 Let us fix α : C R and β : S R and let us introduce the function x H(α, β; x) on Q G by H(α, β; x) = C C det(x C ) α(c) S S det(x S ) ν(s)β(s). Define the measure on Q G by µ G (dx) = H( 1 2 ( C +1), 1 2 ( S +1; x)1 Q G (x)dx. 6

7 Perfect orderings of the cliques. Let C be the family of the k cliques of the connected graph (not necessarily decomposable). Consider a bijection P : {1,..., k} C and S P (j) = [P (1) P (2)... P (j 1)] P (j) for j 2. Then the ordering P is said to be perfect if there exists i j < j such that S P (j) P (i j ). This is a deep notion : a connected graph is decomposable if and only if a perfect ordering of the cliques exists. Furthermore if G is decomposable and if P is perfect then S P (j) is a minimal separator. 7

8 Let us fix a perfect ordering P of the set C of the cliques. For a fixed minimal separator S consider the set of cliques J(P, S) = {C C ; j 2 such that P (j) = C et S P (j) = S}. An important result is that if P is a perfect ordering and if for all S S different from S P (2) one has C J(P,S) (α(c) β(s)) = 0 ( we denote by A P this set of (α, β) s) then by a long calculation one sees that there exists a number Γ(α, β) with the following eigenvalue property : for all y P G Q G e tr xy H(α, β; x)µ G (dx) = Γ(α, β)h(α, β; π(y 1 )). (L.-Massam, Ann. Statist. 2007). 8

9 A reformulation is Q G e tr xψ(x 1) H(α, β; x)µ G (dx) = Γ(α, β)h(α, β; x 1 ) namely the functions x H(α, β; x) are eigenfunctions of the operator f K(f) on functions on Q G defined by K(f)(x 1 ) = Q G e tr xψ(x 1) f(x)µ G (dx). 9

10 This leads to the definition of the Wishart distributions on Q G by 1 Γ(α, β)h(α, β; π(y 1 )) e tr (xy) H(α, β; x)µ G (dx) They are therefore indexed by the shape parameters (α, β) and by the scale parameter y P G. There is an other family of similar formulas where the roles of P G and Q G are exchanged that I have no time to describe. 10

11 Homogeneous graphs and the graph A 4. I have to mention that if G is homogeneous, that is if P G is an homogeneous cone (This happens if and only if G does not contains the chain A 4 : as an induced graph), the above formulas hold for a wider range of parameters α and β than the union of A P where P runs all the perfect orderings. Thus the simplest non homogeneous graph is G = A 4 = with cliques and separators C 1 = {1, 2}, C 2 = {2, 3}, C 3 = {3, 4}, S 2 = {2}, S 3 = {3}. An element of Q G has the form x = x 1 x 12 x 21 x 2 x 23 x 32 x 3 x 34 x 43 x 4 for x Q G, with x ij = x ji, 11

12 Let α i = α(c i ), i = 1, 2, 3 β i = β(s i ), i = 2, 3. Define D = {(α, β) α i > 1 2, i = 1, 2, 3, α 1+α 2 > β 2, α 2 +α 3 > β 3 }. Then the following integral (a 7-uple integral!) converges for all σ Q A4 if and only if (α, β) is in D. Under these conditions, it is equal to e x,ψ(σ) H G (α, β; x)µ G (dx) Q G = Γ(α )Γ(α )Γ(α ) Γ(α 2 ) Γ(α 1 + α 2 β 2 )Γ(α 2 + α 3 β 3 ) π 3 2σ α σα 1+α 2 β σ α 2+α 3 β σ α F 1 (α 1 + α 2 β 2, α 2 + α 3 β 3, α 2, σ2 23 σ 2 σ 3 ) where 2 F 1 denotes the hypergeometric function. NB σ i.j means σ i σ ij σj 1 σ ji, thus line 3 is a function of type H(α, β; σ). 12

13 The two lessons of the example A 4 : The integral has the form CH(α, β; σ) if and only if the hypergeometric function degenerates (we mean when c = a or b for 2 F 1 (a, b; c; x). Therefore (α, β) satisfies the eigenvalue property if and only if it is in the union of the A P s for the 4 perfect orderings P of A The 4 perfect orderings are P 1 = C 1 C 2 C 3, P 2 = C 2 C 1 C 3, but P 3 = C 3 C 2 C 1, P 4 = C 2 C 3 C 1 A P1 = A P2 = {α 2 = β 2 } D A P3 = A P4 = {α 3 = β 3 } D Why? As we are going to see, this is because P 1 and P 2 share the same initial minimal separator, as well as P 4 and P 3. 13

14 What one needs to review about decomposable graphs 1. Junction trees. 2. Minimal separators 3. The two definitions of the multiplicity of a minimal separator. 14

15 Junction trees The cliques of a graph are its maximal complete subsets. A junction tree has the set of cliques as set of vertices and is such that if the clique C is on the unique path from C to C then C C C. For instance is a junction tree for the decomposable graph b m v a where the three cliques are 1 = (amu), 2 = (muvc) and 3 = (bmv). A connected graph is decomposable if and only if a junction tree exists (a neat proof of this is given by Blair and Peyton in 1991) u c 15

16 Minimal separators If a and b are not neighbors S V is a separator of a and b if any path from a to b hits S b m v a For instance muvc is a separator of a and b. If nothing can be taken out, S is a minimal separator of a and b. Finally S is minimal separator by itself if there exist non adjacent a and b such that S is a minimal separator of a and b. There are not so many of them, strictly less that the number of cliques anyway. They are mu and mv in the example. A connected graph is decomposable if and only if all the minimal separators are complete (Dirac 1961). u c 16

17 Topological multiplicity of a minimal separator Let S be a minimal separator of a decomposable graph (V, E). Let {V 1,..., V p } be the connected components of V \ S (of course p 2). Let q be the number of j = 1,..., p such that S is NOT a clique of S V j. The number ν(s) = q 1 is called the topological multiplicity of S. Multiplicity of a minimal separator from a perfect ordering If P is a perfect ordering and if S is a minimal separator, denote by ν P (S) the number of j = 2,..., k such that S = S P (j) ; recall S P (j) = [P (1) P (2)... P (j 1)] P (j). (The topological multiplicity is introduced by Lauritzen, Speed and Vivayan in 1984). Question : one observes that in all cases the two definitions of multiplicity coincide. Why? Answer later on. 17

18 Example : If I remove the minimal separator S = 27 to its graph four connected components are obtained : If I add S to each of them, thus for component 1 I obtain the graph for which S = 27 is a clique. This is not the case for the three other connected components 3, 4 et 56. Therefore q = 3, the topological multiplicity of 27 is 2. 18

19 Similarly since the cliques are C 1 = 12, C 2 = 237, C 3 = 247, C 4 = 2567 one can see that P = C 1 C 2 C 3 C 4 is perfect that S P (3) = S P (4) = 27 and that ν P (27) = 2 = ν(p ). Question : Why do we have always ν P (S) = ν(s)? 19

20 Tiling of a junction tree by the minimal separators If (H, E(H)) is a tree (undirected) with vertex set H and edge set E(H) a tiling of H is a family T of subtrees T = {T 1,..., T p } of H such that if E(T i ) is the edge set of T i then {E(T 1 ),..., E(T q )} is a partition of E(H). This implies T 1... T q = H although (T 1,..., T p ) is not a partition of the set H. 20

21 Example g e c h f d b a the tiles of the tiling can be chosen as g h f e f d b c b b a 21

22 Theorem 1. Let G = (V, E) be a decomposable graph and let (C, E(C)) be a junction tree of G. Let S be the family of minimal separators of G. There exists a unique tiling T of the tree (C, E(C)) by subtrees and a bijection S T S from S towards T with the following property : for all S S the edges of T S are the edges {C, C } such that S = C C. Under these circumstances the number of edges of T S is the topological multiplicity of S. Furthermore if C and C are two distinct cliques consider the unique path (C = C 0, C 1,..., C q = C ) from C to C. Let S i S such that {C i 1, C i } is in T Si. Then C C = q i=1 S i. In particular C C = S if C and C are in T S. 22

23 Consider again the example : There are 4 cliques A = {1, 2}, B = {2, 3, 7}, C = {2, 4, 7}, D = {2, 5, 6, 7} and two minimal separators U = {2}, V = {2, 7}. The ordering ABCD of the cliques is perfect with S 2 = U et S 3 = S 4 = V. Thus V has multiplicity 2 and U has multiplicity 1. Consider the junction tree C A B D Then T U = AB et T V = BCD. 23

24 Junction trees and perfect orderings of cliques. Recall that saying that P is a perfect ordering of the set C of the k cliques of a decomposable graph is to say that there exists i j < j such that S P (j) P (i j ). There exist in general several possible i j s. Actually we fix one such i j for each j and we create the graph having C as vertex set with having the k 1 edges {P (i j ), P (j)}. A beautiful result of Beeri, Fagin, Maier and Yannakakis (1983) claims that this graph is a junction tree and conversely that any junction tree can be constructed from a perfect ordering and from a choice of the j i j. Let us say that a junction tree is adapted to the perfect ordering P if there exists a choice j i j giving the tree. 24

25 Tiling by minimal separators and perfect orderings of the cliques. Let P be a perfect ordering of the set C of the k cliques of a decomposable graph Consider now a junction tree adapted to P and let T be the tiling of this tree by the minimal separators. We transform this undirected tree into a rooted tree by taking P (1) as a root. This transforms C into a partially ordered set : C C if the unique path from P (1) to C passes through C. 26

26 Let S be in the set S of the minimal separators. Recall that we have considered before the set of cliques J(P, S) = {C C ; j 2 such that P (j) = C et S P (j) = S}. Just remark that ν P (S) = J(P, S). Now for all S S the subtree T S has a minimal point M(S) for this partial order. Here is now a useful result ruling out the old contest between multiplicities (recall that the number of vertices of a tree is the number of edges plus one ) : 27

27 Theorem 2. J(P, S) = T S \ {M(S)}. In particular ν P (S) is the topological multiplicity T S 1 of S. 28

28 Actually J(P, S) depends on S and on S P (2) only : Theorem 3. Let P and P two perfect orderings such that P (1) P (2) = P (1) P (2), that is to say S P (2) = S P (2) (denoted S 2 ). Then J(P, S) = J(P, S) if S S 2 and J(P, S 2 ) {P (1)} = J(P, S 2 ) {P (1)}. 29

29 Conclusion : Consequences for the Wishart distributions on decomposable graphs. Recall that given a perfect ordering P, the set A P of acceptable shape parameters (α, β) for the Wishart distribution is the set of (α, β) such that for all minimal separators S we have C J(P,S) (α(c) β(s)) = 0. Thus this crucial set A P depends entirely on the family of subsets of cliques F P = {J(P, S); S S}. This tiling process has shown that actually the family F P -and therefore the set A P of shape parameters - depends only on the first minimal separator S P (2) of the perfect ordering P. 30

MATHEMATICAL ENGINEERING TECHNICAL REPORTS. Boundary cliques, clique trees and perfect sequences of maximal cliques of a chordal graph

MATHEMATICAL ENGINEERING TECHNICAL REPORTS. Boundary cliques, clique trees and perfect sequences of maximal cliques of a chordal graph MATHEMATICAL ENGINEERING TECHNICAL REPORTS Boundary cliques, clique trees and perfect sequences of maximal cliques of a chordal graph Hisayuki HARA and Akimichi TAKEMURA METR 2006 41 July 2006 DEPARTMENT

More information

PIOTR GRACZYK LAREMA, UNIVERSITÉ D ANGERS, FRANCJA. Estymatory kowariancji dla modeli graficznych

PIOTR GRACZYK LAREMA, UNIVERSITÉ D ANGERS, FRANCJA. Estymatory kowariancji dla modeli graficznych PIOTR GRACZYK LAREMA, UNIVERSITÉ D ANGERS, FRANCJA Estymatory kowariancji dla modeli graficznych wspó lautorzy: H. Ishi(Nagoya), B. Ko lodziejek(warszawa), S. Mamane(Johannesburg), H. Ochiai(Fukuoka) XLV

More information

WISHART/RIETZ DISTRIBUTIONS AND DECOMPOSABLE UNDIRECTED GRAPHS BY STEEN ANDERSSON INDIANA UNIVERSITY AND THOMAS KLEIN AUGSBURG UNIVERSITÄET

WISHART/RIETZ DISTRIBUTIONS AND DECOMPOSABLE UNDIRECTED GRAPHS BY STEEN ANDERSSON INDIANA UNIVERSITY AND THOMAS KLEIN AUGSBURG UNIVERSITÄET . WISHART/RIETZ DISTRIBUTIONS AND DECOMPOSABLE UNDIRECTED GRAPHS BY STEEN ANDERSSON INDIANA UNIVERSITY AND THOMAS KLEIN AUGSBURG UNIVERSITÄET 1 TARTU 6JUN007 Classical Wishart distributions. Let V be a

More information

Contingency tables from the algebraic statistics view point Gérard Letac, Université Paul Sabatier, Toulouse

Contingency tables from the algebraic statistics view point Gérard Letac, Université Paul Sabatier, Toulouse Contingency tables from the algebraic statistics view point Gérard Letac, Université Paul Sabatier, Toulouse College Park, October 14 th, 2010.. Joint work with Hélène Massam 1 Contingency tables Let I

More information

PIOTR GRACZYK LAREMA, UNIVERSITÉ D ANGERS, FRANCE. Analysis of Wishart matrices: Riesz and Wishart laws on graphical cones. Letac-Massam conjecture

PIOTR GRACZYK LAREMA, UNIVERSITÉ D ANGERS, FRANCE. Analysis of Wishart matrices: Riesz and Wishart laws on graphical cones. Letac-Massam conjecture PIOTR GRACZYK LAREMA, UNIVERSITÉ D ANGERS, FRANCE Analysis of Wishart matrices: Riesz and Wishart laws on graphical cones Letac-Massam conjecture Joint work with H. Ishi, S. Mamane, H. Ochiai Mathematical

More information

arxiv: v1 [math.st] 16 Aug 2011

arxiv: v1 [math.st] 16 Aug 2011 Retaining positive definiteness in thresholded matrices Dominique Guillot Stanford University Bala Rajaratnam Stanford University August 17, 2011 arxiv:1108.3325v1 [math.st] 16 Aug 2011 Abstract Positive

More information

FLEXIBLE COVARIANCE ESTIMATION IN GRAPHICAL GAUSSIAN MODELS

FLEXIBLE COVARIANCE ESTIMATION IN GRAPHICAL GAUSSIAN MODELS Submitted to the Annals of Statistics FLEXIBLE COVARIANCE ESTIMATION IN GRAPHICAL GAUSSIAN MODELS By Bala Rajaratnam, Hélène Massam and Carlos M. Carvalho Stanford University, York University, The University

More information

The doubly negative matrix completion problem

The doubly negative matrix completion problem The doubly negative matrix completion problem C Mendes Araújo, Juan R Torregrosa and Ana M Urbano CMAT - Centro de Matemática / Dpto de Matemática Aplicada Universidade do Minho / Universidad Politécnica

More information

Lecture 12: May 09, Decomposable Graphs (continues from last time)

Lecture 12: May 09, Decomposable Graphs (continues from last time) 596 Pat. Recog. II: Introduction to Graphical Models University of Washington Spring 00 Dept. of lectrical ngineering Lecture : May 09, 00 Lecturer: Jeff Bilmes Scribe: Hansang ho, Izhak Shafran(000).

More information

Kernels of Directed Graph Laplacians. J. S. Caughman and J.J.P. Veerman

Kernels of Directed Graph Laplacians. J. S. Caughman and J.J.P. Veerman Kernels of Directed Graph Laplacians J. S. Caughman and J.J.P. Veerman Department of Mathematics and Statistics Portland State University PO Box 751, Portland, OR 97207. caughman@pdx.edu, veerman@pdx.edu

More information

FLEXIBLE COVARIANCE ESTIMATION IN GRAPHICAL GAUSSIAN MODELS

FLEXIBLE COVARIANCE ESTIMATION IN GRAPHICAL GAUSSIAN MODELS Submitted to the Annals of Statistics FLEXIBLE COVARIANCE ESTIMATION IN GRAPHICAL GAUSSIAN MODELS By Bala Rajaratnam, Hélène Massam and Carlos M. Carvalho Stanford University, York University, The University

More information

The structure of bull-free graphs II elementary trigraphs

The structure of bull-free graphs II elementary trigraphs The structure of bull-free graphs II elementary trigraphs Maria Chudnovsky Columbia University, New York, NY 10027 USA May 6, 2006; revised April 25, 2011 Abstract The bull is a graph consisting of a triangle

More information

Combinatorial Types of Tropical Eigenvector

Combinatorial Types of Tropical Eigenvector Combinatorial Types of Tropical Eigenvector arxiv:1105.55504 Ngoc Mai Tran Department of Statistics, UC Berkeley Joint work with Bernd Sturmfels 2 / 13 Tropical eigenvalues and eigenvectors Max-plus: (R,,

More information

Semidefinite and Second Order Cone Programming Seminar Fall 2001 Lecture 5

Semidefinite and Second Order Cone Programming Seminar Fall 2001 Lecture 5 Semidefinite and Second Order Cone Programming Seminar Fall 2001 Lecture 5 Instructor: Farid Alizadeh Scribe: Anton Riabov 10/08/2001 1 Overview We continue studying the maximum eigenvalue SDP, and generalize

More information

Clique trees of infinite locally finite chordal graphs

Clique trees of infinite locally finite chordal graphs Clique trees of infinite locally finite chordal graphs Florian Lehner & Christoph Temmel Abstract We investigate clique trees of infinite, locally finite chordal graphs. Our key tool is a bijection between

More information

Graphical Gaussian models and their groups

Graphical Gaussian models and their groups Piotr Zwiernik TU Eindhoven (j.w. Jan Draisma, Sonja Kuhnt) Workshop on Graphical Models, Fields Institute, Toronto, 16 Apr 2012 1 / 23 Outline and references Outline: 1. Invariance of statistical models

More information

ENTANGLED STATES ARISING FROM INDECOMPOSABLE POSITIVE LINEAR MAPS. 1. Introduction

ENTANGLED STATES ARISING FROM INDECOMPOSABLE POSITIVE LINEAR MAPS. 1. Introduction Trends in Mathematics Information Center for Mathematical Sciences Volume 6, Number 2, December, 2003, Pages 83 91 ENTANGLED STATES ARISING FROM INDECOMPOSABLE POSITIVE LINEAR MAPS SEUNG-HYEOK KYE Abstract.

More information

Properties of θ-super positive graphs

Properties of θ-super positive graphs Properties of θ-super positive graphs Cheng Yeaw Ku Department of Mathematics, National University of Singapore, Singapore 117543 matkcy@nus.edu.sg Kok Bin Wong Institute of Mathematical Sciences, University

More information

Partial cubes: structures, characterizations, and constructions

Partial cubes: structures, characterizations, and constructions Partial cubes: structures, characterizations, and constructions Sergei Ovchinnikov San Francisco State University, Mathematics Department, 1600 Holloway Ave., San Francisco, CA 94132 Abstract Partial cubes

More information

The Strong Largeur d Arborescence

The Strong Largeur d Arborescence The Strong Largeur d Arborescence Rik Steenkamp (5887321) November 12, 2013 Master Thesis Supervisor: prof.dr. Monique Laurent Local Supervisor: prof.dr. Alexander Schrijver KdV Institute for Mathematics

More information

Cluster varieties for tree-shaped quivers and their cohomology

Cluster varieties for tree-shaped quivers and their cohomology Cluster varieties for tree-shaped quivers and their cohomology Frédéric Chapoton CNRS & Université de Strasbourg Octobre 2016 Cluster algebras and the associated varieties Cluster algebras are commutative

More information

fy (X(g)) Y (f)x(g) gy (X(f)) Y (g)x(f)) = fx(y (g)) + gx(y (f)) fy (X(g)) gy (X(f))

fy (X(g)) Y (f)x(g) gy (X(f)) Y (g)x(f)) = fx(y (g)) + gx(y (f)) fy (X(g)) gy (X(f)) 1. Basic algebra of vector fields Let V be a finite dimensional vector space over R. Recall that V = {L : V R} is defined to be the set of all linear maps to R. V is isomorphic to V, but there is no canonical

More information

Ma/CS 6b Class 23: Eigenvalues in Regular Graphs

Ma/CS 6b Class 23: Eigenvalues in Regular Graphs Ma/CS 6b Class 3: Eigenvalues in Regular Graphs By Adam Sheffer Recall: The Spectrum of a Graph Consider a graph G = V, E and let A be the adjacency matrix of G. The eigenvalues of G are the eigenvalues

More information

Decomposable Graphical Gaussian Models

Decomposable Graphical Gaussian Models CIMPA Summerschool, Hammamet 2011, Tunisia September 12, 2011 Basic algorithm This simple algorithm has complexity O( V + E ): 1. Choose v 0 V arbitrary and let v 0 = 1; 2. When vertices {1, 2,..., j}

More information

We simply compute: for v = x i e i, bilinearity of B implies that Q B (v) = B(v, v) is given by xi x j B(e i, e j ) =

We simply compute: for v = x i e i, bilinearity of B implies that Q B (v) = B(v, v) is given by xi x j B(e i, e j ) = Math 395. Quadratic spaces over R 1. Algebraic preliminaries Let V be a vector space over a field F. Recall that a quadratic form on V is a map Q : V F such that Q(cv) = c 2 Q(v) for all v V and c F, and

More information

CS281A/Stat241A Lecture 19

CS281A/Stat241A Lecture 19 CS281A/Stat241A Lecture 19 p. 1/4 CS281A/Stat241A Lecture 19 Junction Tree Algorithm Peter Bartlett CS281A/Stat241A Lecture 19 p. 2/4 Announcements My office hours: Tuesday Nov 3 (today), 1-2pm, in 723

More information

Testing Equality of Natural Parameters for Generalized Riesz Distributions

Testing Equality of Natural Parameters for Generalized Riesz Distributions Testing Equality of Natural Parameters for Generalized Riesz Distributions Jesse Crawford Department of Mathematics Tarleton State University jcrawford@tarleton.edu faculty.tarleton.edu/crawford April

More information

Learning from Sensor Data: Set II. Behnaam Aazhang J.S. Abercombie Professor Electrical and Computer Engineering Rice University

Learning from Sensor Data: Set II. Behnaam Aazhang J.S. Abercombie Professor Electrical and Computer Engineering Rice University Learning from Sensor Data: Set II Behnaam Aazhang J.S. Abercombie Professor Electrical and Computer Engineering Rice University 1 6. Data Representation The approach for learning from data Probabilistic

More information

Notes on the Matrix-Tree theorem and Cayley s tree enumerator

Notes on the Matrix-Tree theorem and Cayley s tree enumerator Notes on the Matrix-Tree theorem and Cayley s tree enumerator 1 Cayley s tree enumerator Recall that the degree of a vertex in a tree (or in any graph) is the number of edges emanating from it We will

More information

Graph coloring, perfect graphs

Graph coloring, perfect graphs Lecture 5 (05.04.2013) Graph coloring, perfect graphs Scribe: Tomasz Kociumaka Lecturer: Marcin Pilipczuk 1 Introduction to graph coloring Definition 1. Let G be a simple undirected graph and k a positive

More information

Bayesian & Markov Networks: A unified view

Bayesian & Markov Networks: A unified view School of omputer Science ayesian & Markov Networks: unified view Probabilistic Graphical Models (10-708) Lecture 3, Sep 19, 2007 Receptor Kinase Gene G Receptor X 1 X 2 Kinase Kinase E X 3 X 4 X 5 TF

More information

Fisher Information in Gaussian Graphical Models

Fisher Information in Gaussian Graphical Models Fisher Information in Gaussian Graphical Models Jason K. Johnson September 21, 2006 Abstract This note summarizes various derivations, formulas and computational algorithms relevant to the Fisher information

More information

The non-bipartite graphs with all but two eigenvalues in

The non-bipartite graphs with all but two eigenvalues in The non-bipartite graphs with all but two eigenvalues in [ 1, 1] L.S. de Lima 1, A. Mohammadian 1, C.S. Oliveira 2 1 Departamento de Engenharia de Produção, Centro Federal de Educação Tecnológica Celso

More information

Sparse Matrix Theory and Semidefinite Optimization

Sparse Matrix Theory and Semidefinite Optimization Sparse Matrix Theory and Semidefinite Optimization Lieven Vandenberghe Department of Electrical Engineering University of California, Los Angeles Joint work with Martin S. Andersen and Yifan Sun Third

More information

3 Undirected Graphical Models

3 Undirected Graphical Models Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.438 Algorithms For Inference Fall 2014 3 Undirected Graphical Models In this lecture, we discuss undirected

More information

Claw-free Graphs. III. Sparse decomposition

Claw-free Graphs. III. Sparse decomposition Claw-free Graphs. III. Sparse decomposition Maria Chudnovsky 1 and Paul Seymour Princeton University, Princeton NJ 08544 October 14, 003; revised May 8, 004 1 This research was conducted while the author

More information

A lower bound for the Laplacian eigenvalues of a graph proof of a conjecture by Guo

A lower bound for the Laplacian eigenvalues of a graph proof of a conjecture by Guo A lower bound for the Laplacian eigenvalues of a graph proof of a conjecture by Guo A. E. Brouwer & W. H. Haemers 2008-02-28 Abstract We show that if µ j is the j-th largest Laplacian eigenvalue, and d

More information

Perfect matchings in highly cyclically connected regular graphs

Perfect matchings in highly cyclically connected regular graphs Perfect matchings in highly cyclically connected regular graphs arxiv:1709.08891v1 [math.co] 6 Sep 017 Robert Lukot ka Comenius University, Bratislava lukotka@dcs.fmph.uniba.sk Edita Rollová University

More information

Testing Equality in Communication Graphs

Testing Equality in Communication Graphs Electronic Colloquium on Computational Complexity, Report No. 86 (2016) Testing Equality in Communication Graphs Noga Alon Klim Efremenko Benny Sudakov Abstract Let G = (V, E) be a connected undirected

More information

Introduction to Graphical Models

Introduction to Graphical Models Introduction to Graphical Models STA 345: Multivariate Analysis Department of Statistical Science Duke University, Durham, NC, USA Robert L. Wolpert 1 Conditional Dependence Two real-valued or vector-valued

More information

Undirected Graphical Models 4 Bayesian Networks and Markov Networks. Bayesian Networks to Markov Networks

Undirected Graphical Models 4 Bayesian Networks and Markov Networks. Bayesian Networks to Markov Networks Undirected Graphical Models 4 ayesian Networks and Markov Networks 1 ayesian Networks to Markov Networks 2 1 Ns to MNs X Y Z Ns can represent independence constraints that MN cannot MNs can represent independence

More information

Undirected Graphical Models

Undirected Graphical Models Undirected Graphical Models 1 Conditional Independence Graphs Let G = (V, E) be an undirected graph with vertex set V and edge set E, and let A, B, and C be subsets of vertices. We say that C separates

More information

HOW IS A CHORDAL GRAPH LIKE A SUPERSOLVABLE BINARY MATROID?

HOW IS A CHORDAL GRAPH LIKE A SUPERSOLVABLE BINARY MATROID? HOW IS A CHORDAL GRAPH LIKE A SUPERSOLVABLE BINARY MATROID? RAUL CORDOVIL, DAVID FORGE AND SULAMITA KLEIN To the memory of Claude Berge Abstract. Let G be a finite simple graph. From the pioneering work

More information

Total positivity in Markov structures

Total positivity in Markov structures 1 based on joint work with Shaun Fallat, Kayvan Sadeghi, Caroline Uhler, Nanny Wermuth, and Piotr Zwiernik (arxiv:1510.01290) Faculty of Science Total positivity in Markov structures Steffen Lauritzen

More information

Lecture 7: Semidefinite programming

Lecture 7: Semidefinite programming CS 766/QIC 820 Theory of Quantum Information (Fall 2011) Lecture 7: Semidefinite programming This lecture is on semidefinite programming, which is a powerful technique from both an analytic and computational

More information

A Local Inverse Formula and a Factorization

A Local Inverse Formula and a Factorization A Local Inverse Formula and a Factorization arxiv:60.030v [math.na] 4 Oct 06 Gilbert Strang and Shev MacNamara With congratulations to Ian Sloan! Abstract When a matrix has a banded inverse there is a

More information

TEST CODE: PMB SYLLABUS

TEST CODE: PMB SYLLABUS TEST CODE: PMB SYLLABUS Convergence and divergence of sequence and series; Cauchy sequence and completeness; Bolzano-Weierstrass theorem; continuity, uniform continuity, differentiability; directional

More information

RESEARCH ARTICLE. An extension of the polytope of doubly stochastic matrices

RESEARCH ARTICLE. An extension of the polytope of doubly stochastic matrices Linear and Multilinear Algebra Vol. 00, No. 00, Month 200x, 1 15 RESEARCH ARTICLE An extension of the polytope of doubly stochastic matrices Richard A. Brualdi a and Geir Dahl b a Department of Mathematics,

More information

Rao s degree sequence conjecture

Rao s degree sequence conjecture Rao s degree sequence conjecture Maria Chudnovsky 1 Columbia University, New York, NY 10027 Paul Seymour 2 Princeton University, Princeton, NJ 08544 July 31, 2009; revised December 10, 2013 1 Supported

More information

Gaussian representation of a class of Riesz probability distributions

Gaussian representation of a class of Riesz probability distributions arxiv:763v [mathpr] 8 Dec 7 Gaussian representation of a class of Riesz probability distributions A Hassairi Sfax University Tunisia Running title: Gaussian representation of a Riesz distribution Abstract

More information

Mic ael Flohr Representation theory of semi-simple Lie algebras: Example su(3) 6. and 20. June 2003

Mic ael Flohr Representation theory of semi-simple Lie algebras: Example su(3) 6. and 20. June 2003 Handout V for the course GROUP THEORY IN PHYSICS Mic ael Flohr Representation theory of semi-simple Lie algebras: Example su(3) 6. and 20. June 2003 GENERALIZING THE HIGHEST WEIGHT PROCEDURE FROM su(2)

More information

Claw-Free Graphs With Strongly Perfect Complements. Fractional and Integral Version.

Claw-Free Graphs With Strongly Perfect Complements. Fractional and Integral Version. Claw-Free Graphs With Strongly Perfect Complements. Fractional and Integral Version. Part II. Nontrivial strip-structures Maria Chudnovsky Department of Industrial Engineering and Operations Research Columbia

More information

Real symmetric matrices/1. 1 Eigenvalues and eigenvectors

Real symmetric matrices/1. 1 Eigenvalues and eigenvectors Real symmetric matrices 1 Eigenvalues and eigenvectors We use the convention that vectors are row vectors and matrices act on the right. Let A be a square matrix with entries in a field F; suppose that

More information

U.C. Berkeley Better-than-Worst-Case Analysis Handout 3 Luca Trevisan May 24, 2018

U.C. Berkeley Better-than-Worst-Case Analysis Handout 3 Luca Trevisan May 24, 2018 U.C. Berkeley Better-than-Worst-Case Analysis Handout 3 Luca Trevisan May 24, 2018 Lecture 3 In which we show how to find a planted clique in a random graph. 1 Finding a Planted Clique We will analyze

More information

Lecture 1 and 2: Random Spanning Trees

Lecture 1 and 2: Random Spanning Trees Recent Advances in Approximation Algorithms Spring 2015 Lecture 1 and 2: Random Spanning Trees Lecturer: Shayan Oveis Gharan March 31st Disclaimer: These notes have not been subjected to the usual scrutiny

More information

Characterizations of Strongly Regular Graphs: Part II: Bose-Mesner algebras of graphs. Sung Y. Song Iowa State University

Characterizations of Strongly Regular Graphs: Part II: Bose-Mesner algebras of graphs. Sung Y. Song Iowa State University Characterizations of Strongly Regular Graphs: Part II: Bose-Mesner algebras of graphs Sung Y. Song Iowa State University sysong@iastate.edu Notation K: one of the fields R or C X: a nonempty finite set

More information

Markov Chains and MCMC

Markov Chains and MCMC Markov Chains and MCMC Markov chains Let S = {1, 2,..., N} be a finite set consisting of N states. A Markov chain Y 0, Y 1, Y 2,... is a sequence of random variables, with Y t S for all points in time

More information

Four new upper bounds for the stability number of a graph

Four new upper bounds for the stability number of a graph Four new upper bounds for the stability number of a graph Miklós Ujvári Abstract. In 1979, L. Lovász defined the theta number, a spectral/semidefinite upper bound on the stability number of a graph, which

More information

The random continued fractions of Dyson and their extensions. La Sapientia, January 25th, G. Letac, Université Paul Sabatier, Toulouse.

The random continued fractions of Dyson and their extensions. La Sapientia, January 25th, G. Letac, Université Paul Sabatier, Toulouse. The random continued fractions of Dyson and their extensions La Sapientia, January 25th, 2010. G. Letac, Université Paul Sabatier, Toulouse. 1 56 years ago Freeman J. Dyson the Princeton physicist and

More information

Probabilistic Graphical Models

Probabilistic Graphical Models Probabilistic Graphical Models Brown University CSCI 295-P, Spring 213 Prof. Erik Sudderth Lecture 11: Inference & Learning Overview, Gaussian Graphical Models Some figures courtesy Michael Jordan s draft

More information

MATH 829: Introduction to Data Mining and Analysis Graphical Models I

MATH 829: Introduction to Data Mining and Analysis Graphical Models I MATH 829: Introduction to Data Mining and Analysis Graphical Models I Dominique Guillot Departments of Mathematical Sciences University of Delaware May 2, 2016 1/12 Independence and conditional independence:

More information

Decomposable and Directed Graphical Gaussian Models

Decomposable and Directed Graphical Gaussian Models Decomposable Decomposable and Directed Graphical Gaussian Models Graphical Models and Inference, Lecture 13, Michaelmas Term 2009 November 26, 2009 Decomposable Definition Basic properties Wishart density

More information

The large deviation principle for the Erdős-Rényi random graph

The large deviation principle for the Erdős-Rényi random graph The large deviation principle for the Erdős-Rényi random graph (Courant Institute, NYU) joint work with S. R. S. Varadhan Main objective: how to count graphs with a given property Only consider finite

More information

Random Walks on Graphs. One Concrete Example of a random walk Motivation applications

Random Walks on Graphs. One Concrete Example of a random walk Motivation applications Random Walks on Graphs Outline One Concrete Example of a random walk Motivation applications shuffling cards universal traverse sequence self stabilizing token management scheme random sampling enumeration

More information

UNAVOIDABLE INDUCED SUBGRAPHS IN LARGE GRAPHS WITH NO HOMOGENEOUS SETS

UNAVOIDABLE INDUCED SUBGRAPHS IN LARGE GRAPHS WITH NO HOMOGENEOUS SETS UNAVOIDABLE INDUCED SUBGRAPHS IN LARGE GRAPHS WITH NO HOMOGENEOUS SETS MARIA CHUDNOVSKY, RINGI KIM, SANG-IL OUM, AND PAUL SEYMOUR Abstract. An n-vertex graph is prime if it has no homogeneous set, that

More information

Convex Analysis 2013 Let f : Q R be a strongly convex function with convexity parameter µ>0, where Q R n is a bounded, closed, convex set, which contains the origin. Let Q =conv(q, Q) andconsiderthefunction

More information

A new graph parameter related to bounded rank positive semidefinite matrix completions

A new graph parameter related to bounded rank positive semidefinite matrix completions Mathematical Programming manuscript No. (will be inserted by the editor) Monique Laurent Antonios Varvitsiotis A new graph parameter related to bounded rank positive semidefinite matrix completions the

More information

On fat Hoffman graphs with smallest eigenvalue at least 3

On fat Hoffman graphs with smallest eigenvalue at least 3 On fat Hoffman graphs with smallest eigenvalue at least 3 Qianqian Yang University of Science and Technology of China Ninth Shanghai Conference on Combinatorics Shanghai Jiaotong University May 24-28,

More information

ON MULTI-AVOIDANCE OF RIGHT ANGLED NUMBERED POLYOMINO PATTERNS

ON MULTI-AVOIDANCE OF RIGHT ANGLED NUMBERED POLYOMINO PATTERNS INTEGERS: ELECTRONIC JOURNAL OF COMBINATORIAL NUMBER THEORY 4 (2004), #A21 ON MULTI-AVOIDANCE OF RIGHT ANGLED NUMBERED POLYOMINO PATTERNS Sergey Kitaev Department of Mathematics, University of Kentucky,

More information

Example: multivariate Gaussian Distribution

Example: multivariate Gaussian Distribution School of omputer Science Probabilistic Graphical Models Representation of undirected GM (continued) Eric Xing Lecture 3, September 16, 2009 Reading: KF-chap4 Eric Xing @ MU, 2005-2009 1 Example: multivariate

More information

ON THE NUMBER OF COMPONENTS OF A GRAPH

ON THE NUMBER OF COMPONENTS OF A GRAPH Volume 5, Number 1, Pages 34 58 ISSN 1715-0868 ON THE NUMBER OF COMPONENTS OF A GRAPH HAMZA SI KADDOUR AND ELIAS TAHHAN BITTAR Abstract. Let G := (V, E be a simple graph; for I V we denote by l(i the number

More information

Lecture 5: January 30

Lecture 5: January 30 CS71 Randomness & Computation Spring 018 Instructor: Alistair Sinclair Lecture 5: January 30 Disclaimer: These notes have not been subjected to the usual scrutiny accorded to formal publications. They

More information

1 T 1 = where 1 is the all-ones vector. For the upper bound, let v 1 be the eigenvector corresponding. u:(u,v) E v 1(u)

1 T 1 = where 1 is the all-ones vector. For the upper bound, let v 1 be the eigenvector corresponding. u:(u,v) E v 1(u) CME 305: Discrete Mathematics and Algorithms Instructor: Reza Zadeh (rezab@stanford.edu) Final Review Session 03/20/17 1. Let G = (V, E) be an unweighted, undirected graph. Let λ 1 be the maximum eigenvalue

More information

On the Adjacency Spectra of Hypertrees

On the Adjacency Spectra of Hypertrees On the Adjacency Spectra of Hypertrees arxiv:1711.01466v1 [math.sp] 4 Nov 2017 Gregory J. Clark and Joshua N. Cooper Department of Mathematics University of South Carolina November 7, 2017 Abstract We

More information

Generalized Pigeonhole Properties of Graphs and Oriented Graphs

Generalized Pigeonhole Properties of Graphs and Oriented Graphs Europ. J. Combinatorics (2002) 23, 257 274 doi:10.1006/eujc.2002.0574 Available online at http://www.idealibrary.com on Generalized Pigeonhole Properties of Graphs and Oriented Graphs ANTHONY BONATO, PETER

More information

Antoni Marczyk A NOTE ON ARBITRARILY VERTEX DECOMPOSABLE GRAPHS

Antoni Marczyk A NOTE ON ARBITRARILY VERTEX DECOMPOSABLE GRAPHS Opuscula Mathematica Vol. 6 No. 1 006 Antoni Marczyk A NOTE ON ARBITRARILY VERTEX DECOMPOSABLE GRAPHS Abstract. A graph G of order n is said to be arbitrarily vertex decomposable if for each sequence (n

More information

A Field Extension as a Vector Space

A Field Extension as a Vector Space Chapter 8 A Field Extension as a Vector Space In this chapter, we take a closer look at a finite extension from the point of view that is a vector space over. It is clear, for instance, that any is a linear

More information

Combinatorics for algebraic geometers

Combinatorics for algebraic geometers Combinatorics for algebraic geometers Calculations in enumerative geometry Maria Monks March 17, 214 Motivation Enumerative geometry In the late 18 s, Hermann Schubert investigated problems in what is

More information

Disjoint paths in tournaments

Disjoint paths in tournaments Disjoint paths in tournaments Maria Chudnovsky 1 Columbia University, New York, NY 10027, USA Alex Scott Mathematical Institute, University of Oxford, 24-29 St Giles, Oxford OX1 3LB, UK Paul Seymour 2

More information

7 The structure of graphs excluding a topological minor

7 The structure of graphs excluding a topological minor 7 The structure of graphs excluding a topological minor Grohe and Marx [39] proved the following structure theorem for graphs excluding a topological minor: Theorem 7.1 ([39]). For every positive integer

More information

Introduction to Semidefinite Programming I: Basic properties a

Introduction to Semidefinite Programming I: Basic properties a Introduction to Semidefinite Programming I: Basic properties and variations on the Goemans-Williamson approximation algorithm for max-cut MFO seminar on Semidefinite Programming May 30, 2010 Semidefinite

More information

Reaching a Consensus in a Dynamically Changing Environment A Graphical Approach

Reaching a Consensus in a Dynamically Changing Environment A Graphical Approach Reaching a Consensus in a Dynamically Changing Environment A Graphical Approach M. Cao Yale Univesity A. S. Morse Yale University B. D. O. Anderson Australia National University and National ICT Australia

More information

Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems

Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems Rani M. R, Mohith Jagalmohanan, R. Subashini Binary matrices having simultaneous consecutive

More information

Tree-width and planar minors

Tree-width and planar minors Tree-width and planar minors Alexander Leaf and Paul Seymour 1 Princeton University, Princeton, NJ 08544 May 22, 2012; revised March 18, 2014 1 Supported by ONR grant N00014-10-1-0680 and NSF grant DMS-0901075.

More information

Efficient Reassembling of Graphs, Part 1: The Linear Case

Efficient Reassembling of Graphs, Part 1: The Linear Case Efficient Reassembling of Graphs, Part 1: The Linear Case Assaf Kfoury Boston University Saber Mirzaei Boston University Abstract The reassembling of a simple connected graph G = (V, E) is an abstraction

More information

An Algebraic and Geometric Perspective on Exponential Families

An Algebraic and Geometric Perspective on Exponential Families An Algebraic and Geometric Perspective on Exponential Families Caroline Uhler (IST Austria) Based on two papers: with Mateusz Micha lek, Bernd Sturmfels, and Piotr Zwiernik, and with Liam Solus and Ruriko

More information

MATH 61-02: PRACTICE PROBLEMS FOR FINAL EXAM

MATH 61-02: PRACTICE PROBLEMS FOR FINAL EXAM MATH 61-02: PRACTICE PROBLEMS FOR FINAL EXAM (FP1) The exclusive or operation, denoted by and sometimes known as XOR, is defined so that P Q is true iff P is true or Q is true, but not both. Prove (through

More information

Likelihood Analysis of Gaussian Graphical Models

Likelihood Analysis of Gaussian Graphical Models Faculty of Science Likelihood Analysis of Gaussian Graphical Models Ste en Lauritzen Department of Mathematical Sciences Minikurs TUM 2016 Lecture 2 Slide 1/43 Overview of lectures Lecture 1 Markov Properties

More information

On star forest ascending subgraph decomposition

On star forest ascending subgraph decomposition On star forest ascending subgraph decomposition Josep M. Aroca and Anna Lladó Department of Mathematics, Univ. Politècnica de Catalunya Barcelona, Spain josep.m.aroca@upc.edu,aina.llado@upc.edu Submitted:

More information

Large Cliques and Stable Sets in Undirected Graphs

Large Cliques and Stable Sets in Undirected Graphs Large Cliques and Stable Sets in Undirected Graphs Maria Chudnovsky Columbia University, New York NY 10027 May 4, 2014 Abstract The cochromatic number of a graph G is the minimum number of stable sets

More information

Course 311: Michaelmas Term 2005 Part III: Topics in Commutative Algebra

Course 311: Michaelmas Term 2005 Part III: Topics in Commutative Algebra Course 311: Michaelmas Term 2005 Part III: Topics in Commutative Algebra D. R. Wilkins Contents 3 Topics in Commutative Algebra 2 3.1 Rings and Fields......................... 2 3.2 Ideals...............................

More information

DISCRIMINANTS, SYMMETRIZED GRAPH MONOMIALS, AND SUMS OF SQUARES

DISCRIMINANTS, SYMMETRIZED GRAPH MONOMIALS, AND SUMS OF SQUARES DISCRIMINANTS, SYMMETRIZED GRAPH MONOMIALS, AND SUMS OF SQUARES PER ALEXANDERSSON AND BORIS SHAPIRO Abstract. Motivated by the necessities of the invariant theory of binary forms J. J. Sylvester constructed

More information

CLIQUES IN THE UNION OF GRAPHS

CLIQUES IN THE UNION OF GRAPHS CLIQUES IN THE UNION OF GRAPHS RON AHARONI, ELI BERGER, MARIA CHUDNOVSKY, AND JUBA ZIANI Abstract. Let B and R be two simple graphs with vertex set V, and let G(B, R) be the simple graph with vertex set

More information

Markov and Gibbs Random Fields

Markov and Gibbs Random Fields Markov and Gibbs Random Fields Bruno Galerne bruno.galerne@parisdescartes.fr MAP5, Université Paris Descartes Master MVA Cours Méthodes stochastiques pour l analyse d images Lundi 6 mars 2017 Outline The

More information

Chordal Coxeter Groups

Chordal Coxeter Groups arxiv:math/0607301v1 [math.gr] 12 Jul 2006 Chordal Coxeter Groups John Ratcliffe and Steven Tschantz Mathematics Department, Vanderbilt University, Nashville TN 37240, USA Abstract: A solution of the isomorphism

More information

Lecture 17: May 29, 2002

Lecture 17: May 29, 2002 EE596 Pat. Recog. II: Introduction to Graphical Models University of Washington Spring 2000 Dept. of Electrical Engineering Lecture 17: May 29, 2002 Lecturer: Jeff ilmes Scribe: Kurt Partridge, Salvador

More information

Independencies. Undirected Graphical Models 2: Independencies. Independencies (Markov networks) Independencies (Bayesian Networks)

Independencies. Undirected Graphical Models 2: Independencies. Independencies (Markov networks) Independencies (Bayesian Networks) (Bayesian Networks) Undirected Graphical Models 2: Use d-separation to read off independencies in a Bayesian network Takes a bit of effort! 1 2 (Markov networks) Use separation to determine independencies

More information

Copositive Plus Matrices

Copositive Plus Matrices Copositive Plus Matrices Willemieke van Vliet Master Thesis in Applied Mathematics October 2011 Copositive Plus Matrices Summary In this report we discuss the set of copositive plus matrices and their

More information

Markov Random Fields

Markov Random Fields Markov Random Fields 1. Markov property The Markov property of a stochastic sequence {X n } n 0 implies that for all n 1, X n is independent of (X k : k / {n 1, n, n + 1}), given (X n 1, X n+1 ). Another

More information

Classification of semisimple Lie algebras

Classification of semisimple Lie algebras Chapter 6 Classification of semisimple Lie algebras When we studied sl 2 (C), we discovered that it is spanned by elements e, f and h fulfilling the relations: [e, h] = 2e, [ f, h] = 2 f and [e, f ] =

More information