A Statistical Look at Spectral Graph Analysis. Deep Mukhopadhyay
|
|
- Madlyn Cameron
- 5 years ago
- Views:
Transcription
1 A Statistical Look at Spectral Graph Analysis Deep Mukhopadhyay Department of Statistics, Temple University Office: Speakman Graph Signal Processing Workshop, Philadelphia University of Pennsylvania: May 27, 2016 Deep Statistical Spectral Graph Analysis May / 15
2 1. Graph Signal Processing: A Practitioner s Guide 1 Graph Signal Processing = Spectral graph theory + Harmonic analysis. 2 Choose basis as eigenfunction of S. Common choices of S (1) L = D 1/2 AD 1/2 ; Fiedler (1973) (2) R = D 1 A Coifman, and Lafon. (2006) (3) B = A N 1 dd T ; Newman, M. E. (2006) (4) Type-I Reg. L = D 1/2 τ A D 1/2 τ ; Chaudhuri et al. (2012) (5) Type-II Reg. L = D 1/2 τ A τ D 1/2 τ Amini et al. (2013) D = diag(d 1,..., d n ) R n n, where d i denotes the degree of a node. 3 Compute Graph Fourier Transform by expanding signals (defined over the vertices of a graph) as linear combination of the selected eigenbasis and carry out learning tasks such as regression, clustering, classification, smoothing, kriging, etc. in a straightforward manner.
3 2. Can Beauty and Utility Coexist? The way in which spectral graph analysis is currently taught and practiced is rather mechanical, consisting of a series of matrix calculations this has a huge negative bearing on our understanding. How can we make the approach less mechanistic and more systematic ( Scientific)? What unifying feature is shared by all spectral graph analysis approaches based on different shift operators? How can we establish a statistical path to discover these different spectral methods based on few general principles? Most of these questions are either unsolved or unasked. The main challenge to discover the right starting point (NOT the end products). Where should we start? Can we develop a systematic constructive theory starting from that fundamental object from scratch?
4 3. Unified Construction Principle Step 1. For given discrete graph G of size n, construct GraField kernel function C : [0, 1] 2 R + {0} defined a.e by C (u, v; G n ) = p( ) Q(u; X ), Q(v; Y ); G n p ( Q(u; X ) ) p ( ), 0 < u, v < 1, (1) Q(v; Y ) where u = F (x; X ), v = F (y; Y ) for x, y {1, 2,..., n} and degree sequence induced graph mass functions n n p(x; X ) = A(x, y)/n, p(y; Y ) = A(x, y)/n, p(x, y; G) = A(x, y)/n. y=1 x=1 with Q(u; X ) and Q(v; Y ) are the respective quantile functions. Step 2. ξ j (X ; F (X )) denotes polynomials rank-transform that are an orthonormal basis for L 2 (F ): E F [ξ j (X ; F (X ))] = 0, and E F [ξ j (X ; F (X ))ξ k (X ; F (X ))] = δ jk.
5 Step 3. Transform Coding of Graphs. Construct generalized graph matrix M(G, ξ) R n n with respect to an orthonormal system ξ: M[j, k; G, ξ] = ξ j, 1 0 (C 1)ξ k L 2 [0,1] for j, k = 1,..., n. (2) They can be viewed as a coefficient matrix of the orthogonal series expansion of C (u, v; G) with respect to the product bases {ξ j ξ k } 1 j,k n. Step 4. Perform the singular value decomposition (SVD) of M = UΛU T = k u kµ k u T k, where where u ij are the elements of the singular vector of moment matrix U = (u 1,..., u n ), and Λ = diag(µ 1,..., µ n ), µ 1 µ n 0. Step 5. Obtain approximate Karhunen-Loéve (KL) representation basis (which act as a Fourier basis) of the graph G by φ k (u) = n u jk ξ j, for k = 1,..., n 1. j=1
6 4. GraField: Some Insights and Properties 1. GraField is a positive piecewise-constant kernel satisfying C (u, v; G) du dv = C (u, v; G) du dv = 1, [0,1] 2 I ij (i,j) {1,...,n} 2 where { 1, if (u, v) (F (i; X ), F (i + 1; X )] (F (j; Y ), F (j + 1; Y )] I ij (u, v) = 0, elsewhere. 2. In the continuum limit (as the dimension of the graph n ), the shape of the piecewise-constant discrete C approaches to a continuous field over unit interval. 3. The slices of the GraField kernel can be expressed p(y x; G)/p(y; G) in the vertex domain, suggesting a connection with the random walk on the graph. Interpret p(y x; G) as transition probability from vertex x to vertex y, and (when G is non-bipartite) p(y; G) as the stationary distribution.
7 5. Karhunen-Loéve (KL) Representation of Graph We define the Karhunen-Loéve (KL) representation of a graph G based on spectral expansion of its GCD function C (u, v; G). The pioneering work by Schmidt (1907) guarantee the existence of the following decomposition result for undirected graph. Theorem 1. The L 2 GCD bivariate kernel C : [0, 1] 2 R + {0} admits the following canonical representation C (u, v; G n ) = 1 + n λ k φ k (u)φ k (v), (3) k=1 where the non-negative λ 1 λ 2 λ n 0 are singular values and {φ k } k 1 are the orthonormal singular functions φ j, φ k L 2 [0,1] = δ jk, for j, k = 1,..., n, which can be evaluate as the solution of the following integral equation relation [C (u, v; G) 1] φ k (v) dv = λ k φ k (u), k = 1, 2,..., n. (4) [0,1]
8 6. Nonparametric Spectral Approximation Definition 3. The FUNDAMENTAL statistical modeling problem. We define Spectral graph learning algorithm as method of approximating the singular system (λ k, φ k ) k 1 that satisfies the integral equation (4). { A ( λ1, φ ) 1,..., ( λn, φ ) } n Definition 4. Orthogonal series spectral approximation (SOS): Approximate the unknown function φ k as a linear combination of elements from a complete orthogonal system in L 2 [0, 1]. Let {ξ k } be a complete basis of R n defined on the unit interval [0, 1]. Accordingly, each singular functions φ k can be expressed as the expansion over this basis φ k (u) = j α jk ξ j (u), u [0, 1]. (5) where α jk are the unknown coefficients to be determined. Deep (Statistics@Fox) Statistical Spectral Graph Analysis May / 15
9 7. Connection with Laplacian Spectral Analysis Degree-Adaptive Block-pulse Basis Functions. Define block-pulse basis functions, in short BPFs) on the non-uniform mesh 0 = u 0 < u 1 < u n = 1 over [0,1], where u j = x j p(x; X ) with local support { p ξ j (u) = 1/2 (j) for u j 1 < u u j ; (6) 0 elsewhere. They are disjoint, orthogonal and complete set of functions satisfying 1 0 ξ j (u) du = p(j), 1 0 ξ 2 j (u) du = 1, and 1 0 ξ j (u)ξ k (u) du = δ jk. Theorem 2. Then the solution of the integral equation (4) for block-pulse orthogonal series approximated (6) Fourier coefficients {α jk } can equivalently be written down in closed form as the following matrix eigen-value problem L [α] = λα, (7) where L = L uu T, L is the Laplacian matrix, and u = D 1/2 p 1 n.
10 8. Connection with Diffusion Map Theorem 3. The empirical GraField admits the following vertex-domain spectral decomposition p(y x; G) p(y; G) = 1 + k λ k φ k (x) φ k (y), (8) where φ k = Dp 1/2 u k, (u k is the kth eigenvector of the Laplacian matrix L), (φ k F )( ) is abbreviated as φ k ( ), p(y x; G) = T (x, y), and T = D 1 A is the transition matrix of a random walk on G with stationary distribution p(y; G) = d y /N. NOTE: Since { φ k } n 1 k=1 approximate the optimal Karhunen-Loeve representation basis, it is only natural to use them for non-linear embedding of graphs. Our approach provides an additional insight and justification for the diffusion coordinates by interpreting it as the strength of connectivity profile for each vertex.
11 9. Connection with Modularity Spectral Analysis Theorem 4. To approximate KL graph basis φ k = j α jkξ j, choose ξ j (u) = I(u j 1 < u u j ) to be characteristic function satisfying 1 0 ξ j (u) du = 1 0 ξ 2 j (u) du = p(j; G). Then the corresponding spectral estimating equation can equivalently be reduced to the following generalized eigenvalue equation in terms of the matrix B = A N 1 dd T Bα = λdα. (9) NOTE: The matrix B, known as modularity matrix, was introduced by Newman (2006) from an entirely different motivation.
12 Spectral Regularization and High-d Discrete Parameter Space 1 Recall that the amplitude of the top hat indicator basis functions {ξ k } depends on the unknown distribution p(x; G) for x = [n]. 2 MLE estimate the unknown distribution (p 1, p 2,..., p n ) (support size = size of the graph = n) is known to be extremely inefficient for N/n = O(1) Large sparse graph where both n and N are of comparable order. 3 A simple solution which is remarkably serviceable is the Laplace/Additive smoothing Raw-empirical MLE estimates: p(j; G) = d j N ; Smooth Laplace estimates: p(j; G) = d j + τ N + nτ { 1 Laplace estimator; τ = 1/2 Krichevsky Trofimov estimator (j = 1,..., n).
13 10. Connection with Type-I Reg. Laplacian τ-regularized indicator basis. Construct τ-regularized top hat indicator basis ξ j;τ by replacing the amplitude p 1/2 (j) by pτ 1/2 (j). Theorem 5. τ-regularized block-pulse series based spectral approximation scheme is equivalent to representing or embedding discrete graphs in the continuous eigenspace of Type-I Regularized Laplacian = D 1/2 τ A D 1/2 τ, (10) where D τ is a diagonal matrix with i-th entry d i + τ. More Surprise: This EXACT regularized Laplacian formula was proposed by Chaudhuri et al. (2012) and Qin and Rohe (2013).
14 11. Connection with Type-II Reg. Laplacian Theorem 6. Estimate the joint probability p(j, k; G) by extending the univariate formula for two-dimensional case as follows: N p τ (j, k; G) = p(j, k; G) + nτ ( ) 1 N + nτ N + nτ n 2, (11) which is equivalent to replacing the original adjacency matrix by A τ = A + (τ/n)11 T. Note that this strategy automatically produces the Laplace-smoothed marginals. Verify this leads to the following spectral graph matrix Type-II Regularized Laplacian = D 1/2 τ A τ D 1/2 τ. (12) Surprise... IDENTICAL to the proposal given in Amini et al. (2013), thus provides a more FUNDAMENTAL understanding of the spectral regularization (consequence: HD Discrete data smoothing), which previously considered as empirical guesswork based fine-tuned solution.
15 Final Remark: Nonparametric Spectral Analysis of Graphs The Gist Spectral Graph Analysis can be transformed into the following canonical graph learning problem: A method of obtaining an approximate Karhunen-Loéve basis functions of GraField C (u, v; G) via orthogonal series expansion, by solving Graph Co-moment based estimating equation. This technique serves as a general-purpose unified framework for understanding the spectral graph theory from statistical perspective. Our formalism extract ALL the pieces of SGT in a coherent manner (by revealing the underlying connections), which have been discovered by many researchers using different approaches and reasoning. Our statistical viewpoint allows generalization and inspire to develop new computational methods that could be necessary to handle large-scale discrete graph problems. Thanks.
Unified Statistical Theory of Spectral Graph Analysis
Unified Statistical Theory of Spectral Graph Analysis Subhadeep Mukhopadhyay Department of Statistical Science, Temple University Philadelphia, Pennsylvania, 19122, U.S.A. Dedicated to the beloved memory
More informationarxiv: v4 [math.st] 20 Sep 2016
Unified Statistical Theory of Spectral Graph Analysis Subhadeep Mukhopadhyay Department of Statistical Science, Temple University Philadelphia, Pennsylvania, 19122, U.S.A. Dedicated to the beloved memory
More informationData Analysis and Manifold Learning Lecture 3: Graphs, Graph Matrices, and Graph Embeddings
Data Analysis and Manifold Learning Lecture 3: Graphs, Graph Matrices, and Graph Embeddings Radu Horaud INRIA Grenoble Rhone-Alpes, France Radu.Horaud@inrialpes.fr http://perception.inrialpes.fr/ Outline
More informationSpectral Algorithms I. Slides based on Spectral Mesh Processing Siggraph 2010 course
Spectral Algorithms I Slides based on Spectral Mesh Processing Siggraph 2010 course Why Spectral? A different way to look at functions on a domain Why Spectral? Better representations lead to simpler solutions
More informationSpectral Theory of Unsigned and Signed Graphs Applications to Graph Clustering. Some Slides
Spectral Theory of Unsigned and Signed Graphs Applications to Graph Clustering Some Slides Jean Gallier Department of Computer and Information Science University of Pennsylvania Philadelphia, PA 19104,
More information1 Matrix notation and preliminaries from spectral graph theory
Graph clustering (or community detection or graph partitioning) is one of the most studied problems in network analysis. One reason for this is that there are a variety of ways to define a cluster or community.
More information17.1 Directed Graphs, Undirected Graphs, Incidence Matrices, Adjacency Matrices, Weighted Graphs
Chapter 17 Graphs and Graph Laplacians 17.1 Directed Graphs, Undirected Graphs, Incidence Matrices, Adjacency Matrices, Weighted Graphs Definition 17.1. A directed graph isa pairg =(V,E), where V = {v
More informationCertifying the Global Optimality of Graph Cuts via Semidefinite Programming: A Theoretic Guarantee for Spectral Clustering
Certifying the Global Optimality of Graph Cuts via Semidefinite Programming: A Theoretic Guarantee for Spectral Clustering Shuyang Ling Courant Institute of Mathematical Sciences, NYU Aug 13, 2018 Joint
More informationChapter 7: Symmetric Matrices and Quadratic Forms
Chapter 7: Symmetric Matrices and Quadratic Forms (Last Updated: December, 06) These notes are derived primarily from Linear Algebra and its applications by David Lay (4ed). A few theorems have been moved
More information(a) If A is a 3 by 4 matrix, what does this tell us about its nullspace? Solution: dim N(A) 1, since rank(a) 3. Ax =
. (5 points) (a) If A is a 3 by 4 matrix, what does this tell us about its nullspace? dim N(A), since rank(a) 3. (b) If we also know that Ax = has no solution, what do we know about the rank of A? C(A)
More informationSpectral Clustering. Guokun Lai 2016/10
Spectral Clustering Guokun Lai 2016/10 1 / 37 Organization Graph Cut Fundamental Limitations of Spectral Clustering Ng 2002 paper (if we have time) 2 / 37 Notation We define a undirected weighted graph
More information8. Diagonalization.
8. Diagonalization 8.1. Matrix Representations of Linear Transformations Matrix of A Linear Operator with Respect to A Basis We know that every linear transformation T: R n R m has an associated standard
More informationGraph Partitioning Using Random Walks
Graph Partitioning Using Random Walks A Convex Optimization Perspective Lorenzo Orecchia Computer Science Why Spectral Algorithms for Graph Problems in practice? Simple to implement Can exploit very efficient
More informationTHE HIDDEN CONVEXITY OF SPECTRAL CLUSTERING
THE HIDDEN CONVEXITY OF SPECTRAL CLUSTERING Luis Rademacher, Ohio State University, Computer Science and Engineering. Joint work with Mikhail Belkin and James Voss This talk A new approach to multi-way
More informationDiffusion and random walks on graphs
Diffusion and random walks on graphs Leonid E. Zhukov School of Data Analysis and Artificial Intelligence Department of Computer Science National Research University Higher School of Economics Structural
More informationData Analysis and Manifold Learning Lecture 7: Spectral Clustering
Data Analysis and Manifold Learning Lecture 7: Spectral Clustering Radu Horaud INRIA Grenoble Rhone-Alpes, France Radu.Horaud@inrialpes.fr http://perception.inrialpes.fr/ Outline of Lecture 7 What is spectral
More informationLab 8: Measuring Graph Centrality - PageRank. Monday, November 5 CompSci 531, Fall 2018
Lab 8: Measuring Graph Centrality - PageRank Monday, November 5 CompSci 531, Fall 2018 Outline Measuring Graph Centrality: Motivation Random Walks, Markov Chains, and Stationarity Distributions Google
More informationLatent Semantic Analysis. Hongning Wang
Latent Semantic Analysis Hongning Wang CS@UVa VS model in practice Document and query are represented by term vectors Terms are not necessarily orthogonal to each other Synonymy: car v.s. automobile Polysemy:
More informationSpectral Clustering. by HU Pili. June 16, 2013
Spectral Clustering by HU Pili June 16, 2013 Outline Clustering Problem Spectral Clustering Demo Preliminaries Clustering: K-means Algorithm Dimensionality Reduction: PCA, KPCA. Spectral Clustering Framework
More informationGraphs, Geometry and Semi-supervised Learning
Graphs, Geometry and Semi-supervised Learning Mikhail Belkin The Ohio State University, Dept of Computer Science and Engineering and Dept of Statistics Collaborators: Partha Niyogi, Vikas Sindhwani In
More informationSpectral Processing. Misha Kazhdan
Spectral Processing Misha Kazhdan [Taubin, 1995] A Signal Processing Approach to Fair Surface Design [Desbrun, et al., 1999] Implicit Fairing of Arbitrary Meshes [Vallet and Levy, 2008] Spectral Geometry
More informationORIE 6334 Spectral Graph Theory September 22, Lecture 11
ORIE 6334 Spectral Graph Theory September, 06 Lecturer: David P. Williamson Lecture Scribe: Pu Yang In today s lecture we will focus on discrete time random walks on undirected graphs. Specifically, we
More informationSpectral Graph Theory and its Applications. Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity
Spectral Graph Theory and its Applications Daniel A. Spielman Dept. of Computer Science Program in Applied Mathematics Yale Unviersity Outline Adjacency matrix and Laplacian Intuition, spectral graph drawing
More informationSpectral Analysis of k-balanced Signed Graphs
Spectral Analysis of k-balanced Signed Graphs Leting Wu 1, Xiaowei Ying 1, Xintao Wu 1, Aidong Lu 1 and Zhi-Hua Zhou 2 1 University of North Carolina at Charlotte, USA, {lwu8,xying, xwu,alu1}@uncc.edu
More informationACM/CMS 107 Linear Analysis & Applications Fall 2017 Assignment 2: PDEs and Finite Element Methods Due: 7th November 2017
ACM/CMS 17 Linear Analysis & Applications Fall 217 Assignment 2: PDEs and Finite Element Methods Due: 7th November 217 For this assignment the following MATLAB code will be required: Introduction http://wwwmdunloporg/cms17/assignment2zip
More informationFinding normalized and modularity cuts by spectral clustering. Ljubjana 2010, October
Finding normalized and modularity cuts by spectral clustering Marianna Bolla Institute of Mathematics Budapest University of Technology and Economics marib@math.bme.hu Ljubjana 2010, October Outline Find
More informationJordan normal form notes (version date: 11/21/07)
Jordan normal form notes (version date: /2/7) If A has an eigenbasis {u,, u n }, ie a basis made up of eigenvectors, so that Au j = λ j u j, then A is diagonal with respect to that basis To see this, let
More information1 Matrix notation and preliminaries from spectral graph theory
Graph clustering (or community detection or graph partitioning) is one of the most studied problems in network analysis. One reason for this is that there are a variety of ways to define a cluster or community.
More informationSpectral Clustering. Spectral Clustering? Two Moons Data. Spectral Clustering Algorithm: Bipartioning. Spectral methods
Spectral Clustering Seungjin Choi Department of Computer Science POSTECH, Korea seungjin@postech.ac.kr 1 Spectral methods Spectral Clustering? Methods using eigenvectors of some matrices Involve eigen-decomposition
More informationMA 575 Linear Models: Cedric E. Ginestet, Boston University Regularization: Ridge Regression and Lasso Week 14, Lecture 2
MA 575 Linear Models: Cedric E. Ginestet, Boston University Regularization: Ridge Regression and Lasso Week 14, Lecture 2 1 Ridge Regression Ridge regression and the Lasso are two forms of regularized
More informationBeyond Scalar Affinities for Network Analysis or Vector Diffusion Maps and the Connection Laplacian
Beyond Scalar Affinities for Network Analysis or Vector Diffusion Maps and the Connection Laplacian Amit Singer Princeton University Department of Mathematics and Program in Applied and Computational Mathematics
More informationPredicting Graph Labels using Perceptron. Shuang Song
Predicting Graph Labels using Perceptron Shuang Song shs037@eng.ucsd.edu Online learning over graphs M. Herbster, M. Pontil, and L. Wainer, Proc. 22nd Int. Conf. Machine Learning (ICML'05), 2005 Prediction
More informationMarch 13, Paper: R.R. Coifman, S. Lafon, Diffusion maps ([Coifman06]) Seminar: Learning with Graphs, Prof. Hein, Saarland University
Kernels March 13, 2008 Paper: R.R. Coifman, S. Lafon, maps ([Coifman06]) Seminar: Learning with Graphs, Prof. Hein, Saarland University Kernels Figure: Example Application from [LafonWWW] meaningful geometric
More informationMarkov Chains, Random Walks on Graphs, and the Laplacian
Markov Chains, Random Walks on Graphs, and the Laplacian CMPSCI 791BB: Advanced ML Sridhar Mahadevan Random Walks! There is significant interest in the problem of random walks! Markov chain analysis! Computer
More informationLinear Algebra: Matrix Eigenvalue Problems
CHAPTER8 Linear Algebra: Matrix Eigenvalue Problems Chapter 8 p1 A matrix eigenvalue problem considers the vector equation (1) Ax = λx. 8.0 Linear Algebra: Matrix Eigenvalue Problems Here A is a given
More informationCOMPSCI 514: Algorithms for Data Science
COMPSCI 514: Algorithms for Data Science Arya Mazumdar University of Massachusetts at Amherst Fall 2018 Lecture 8 Spectral Clustering Spectral clustering Curse of dimensionality Dimensionality Reduction
More informationLecture 1: From Data to Graphs, Weighted Graphs and Graph Laplacian
Lecture 1: From Data to Graphs, Weighted Graphs and Graph Laplacian Radu Balan February 5, 2018 Datasets diversity: Social Networks: Set of individuals ( agents, actors ) interacting with each other (e.g.,
More informationMarkov Chains and Spectral Clustering
Markov Chains and Spectral Clustering Ning Liu 1,2 and William J. Stewart 1,3 1 Department of Computer Science North Carolina State University, Raleigh, NC 27695-8206, USA. 2 nliu@ncsu.edu, 3 billy@ncsu.edu
More informationCS224W: Social and Information Network Analysis Jure Leskovec, Stanford University
CS224W: Social and Information Network Analysis Jure Leskovec Stanford University Jure Leskovec, Stanford University http://cs224w.stanford.edu Task: Find coalitions in signed networks Incentives: European
More informationGAUSSIAN PROCESS TRANSFORMS
GAUSSIAN PROCESS TRANSFORMS Philip A. Chou Ricardo L. de Queiroz Microsoft Research, Redmond, WA, USA pachou@microsoft.com) Computer Science Department, Universidade de Brasilia, Brasilia, Brazil queiroz@ieee.org)
More informationNetwork Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec
Network Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec Jiezhong Qiu Tsinghua University February 21, 2018 Joint work with Yuxiao Dong (MSR), Hao Ma (MSR), Jian Li (IIIS,
More information1. The Polar Decomposition
A PERSONAL INTERVIEW WITH THE SINGULAR VALUE DECOMPOSITION MATAN GAVISH Part. Theory. The Polar Decomposition In what follows, F denotes either R or C. The vector space F n is an inner product space with
More information1 Last time: least-squares problems
MATH Linear algebra (Fall 07) Lecture Last time: least-squares problems Definition. If A is an m n matrix and b R m, then a least-squares solution to the linear system Ax = b is a vector x R n such that
More informationMath 307 Learning Goals. March 23, 2010
Math 307 Learning Goals March 23, 2010 Course Description The course presents core concepts of linear algebra by focusing on applications in Science and Engineering. Examples of applications from recent
More informationLatent Semantic Analysis. Hongning Wang
Latent Semantic Analysis Hongning Wang CS@UVa Recap: vector space model Represent both doc and query by concept vectors Each concept defines one dimension K concepts define a high-dimensional space Element
More informationLecture 12 : Graph Laplacians and Cheeger s Inequality
CPS290: Algorithmic Foundations of Data Science March 7, 2017 Lecture 12 : Graph Laplacians and Cheeger s Inequality Lecturer: Kamesh Munagala Scribe: Kamesh Munagala Graph Laplacian Maybe the most beautiful
More informationRandom Sampling of Bandlimited Signals on Graphs
Random Sampling of Bandlimited Signals on Graphs Pierre Vandergheynst École Polytechnique Fédérale de Lausanne (EPFL) School of Engineering & School of Computer and Communication Sciences Joint work with
More informationData Analysis and Manifold Learning Lecture 9: Diffusion on Manifolds and on Graphs
Data Analysis and Manifold Learning Lecture 9: Diffusion on Manifolds and on Graphs Radu Horaud INRIA Grenoble Rhone-Alpes, France Radu.Horaud@inrialpes.fr http://perception.inrialpes.fr/ Outline of Lecture
More information1 Data Arrays and Decompositions
1 Data Arrays and Decompositions 1.1 Variance Matrices and Eigenstructure Consider a p p positive definite and symmetric matrix V - a model parameter or a sample variance matrix. The eigenstructure is
More informationSpectral Graph Theory
Spectral Graph Theory Aaron Mishtal April 27, 2016 1 / 36 Outline Overview Linear Algebra Primer History Theory Applications Open Problems Homework Problems References 2 / 36 Outline Overview Linear Algebra
More informationHilbert Space Methods for Reduced-Rank Gaussian Process Regression
Hilbert Space Methods for Reduced-Rank Gaussian Process Regression Arno Solin and Simo Särkkä Aalto University, Finland Workshop on Gaussian Process Approximation Copenhagen, Denmark, May 2015 Solin &
More informationEstimating network degree distributions from sampled networks: An inverse problem
Estimating network degree distributions from sampled networks: An inverse problem Eric D. Kolaczyk Dept of Mathematics and Statistics, Boston University kolaczyk@bu.edu Introduction: Networks and Degree
More information5 Linear Algebra and Inverse Problem
5 Linear Algebra and Inverse Problem 5.1 Introduction Direct problem ( Forward problem) is to find field quantities satisfying Governing equations, Boundary conditions, Initial conditions. The direct problem
More informationCS168: The Modern Algorithmic Toolbox Lectures #11 and #12: Spectral Graph Theory
CS168: The Modern Algorithmic Toolbox Lectures #11 and #12: Spectral Graph Theory Tim Roughgarden & Gregory Valiant May 2, 2016 Spectral graph theory is the powerful and beautiful theory that arises from
More informationData dependent operators for the spatial-spectral fusion problem
Data dependent operators for the spatial-spectral fusion problem Wien, December 3, 2012 Joint work with: University of Maryland: J. J. Benedetto, J. A. Dobrosotskaya, T. Doster, K. W. Duke, M. Ehler, A.
More information22m:033 Notes: 7.1 Diagonalization of Symmetric Matrices
m:33 Notes: 7. Diagonalization of Symmetric Matrices Dennis Roseman University of Iowa Iowa City, IA http://www.math.uiowa.edu/ roseman May 3, Symmetric matrices Definition. A symmetric matrix is a matrix
More informationSpectra of Adjacency and Laplacian Matrices
Spectra of Adjacency and Laplacian Matrices Definition: University of Alicante (Spain) Matrix Computing (subject 3168 Degree in Maths) 30 hours (theory)) + 15 hours (practical assignment) Contents 1. Spectra
More informationQuadratic forms. Here. Thus symmetric matrices are diagonalizable, and the diagonalization can be performed by means of an orthogonal matrix.
Quadratic forms 1. Symmetric matrices An n n matrix (a ij ) n ij=1 with entries on R is called symmetric if A T, that is, if a ij = a ji for all 1 i, j n. We denote by S n (R) the set of all n n symmetric
More informationLearning gradients: prescriptive models
Department of Statistical Science Institute for Genome Sciences & Policy Department of Computer Science Duke University May 11, 2007 Relevant papers Learning Coordinate Covariances via Gradients. Sayan
More informationChapter 11. Matrix Algorithms and Graph Partitioning. M. E. J. Newman. June 10, M. E. J. Newman Chapter 11 June 10, / 43
Chapter 11 Matrix Algorithms and Graph Partitioning M. E. J. Newman June 10, 2016 M. E. J. Newman Chapter 11 June 10, 2016 1 / 43 Table of Contents 1 Eigenvalue and Eigenvector Eigenvector Centrality The
More informationUnsupervised Learning Techniques Class 07, 1 March 2006 Andrea Caponnetto
Unsupervised Learning Techniques 9.520 Class 07, 1 March 2006 Andrea Caponnetto About this class Goal To introduce some methods for unsupervised learning: Gaussian Mixtures, K-Means, ISOMAP, HLLE, Laplacian
More informationKarhunen-Loève Approximation of Random Fields Using Hierarchical Matrix Techniques
Institut für Numerische Mathematik und Optimierung Karhunen-Loève Approximation of Random Fields Using Hierarchical Matrix Techniques Oliver Ernst Computational Methods with Applications Harrachov, CR,
More informationLecture 13 Spectral Graph Algorithms
COMS 995-3: Advanced Algorithms March 6, 7 Lecture 3 Spectral Graph Algorithms Instructor: Alex Andoni Scribe: Srikar Varadaraj Introduction Today s topics: Finish proof from last lecture Example of random
More informationReview problems for MA 54, Fall 2004.
Review problems for MA 54, Fall 2004. Below are the review problems for the final. They are mostly homework problems, or very similar. If you are comfortable doing these problems, you should be fine on
More informationSummary of Week 9 B = then A A =
Summary of Week 9 Finding the square root of a positive operator Last time we saw that positive operators have a unique positive square root We now briefly look at how one would go about calculating the
More informationNotes on singular value decomposition for Math 54. Recall that if A is a symmetric n n matrix, then A has real eigenvalues A = P DP 1 A = P DP T.
Notes on singular value decomposition for Math 54 Recall that if A is a symmetric n n matrix, then A has real eigenvalues λ 1,, λ n (possibly repeated), and R n has an orthonormal basis v 1,, v n, where
More informationORIE 4741: Learning with Big Messy Data. Spectral Graph Theory
ORIE 4741: Learning with Big Messy Data Spectral Graph Theory Mika Sumida Operations Research and Information Engineering Cornell September 15, 2017 1 / 32 Outline Graph Theory Spectral Graph Theory Laplacian
More informationLinear Algebra using Dirac Notation: Pt. 2
Linear Algebra using Dirac Notation: Pt. 2 PHYS 476Q - Southern Illinois University February 6, 2018 PHYS 476Q - Southern Illinois University Linear Algebra using Dirac Notation: Pt. 2 February 6, 2018
More informationNetworks and Their Spectra
Networks and Their Spectra Victor Amelkin University of California, Santa Barbara Department of Computer Science victor@cs.ucsb.edu December 4, 2017 1 / 18 Introduction Networks (= graphs) are everywhere.
More informationData Mining and Analysis: Fundamental Concepts and Algorithms
Data Mining and Analysis: Fundamental Concepts and Algorithms dataminingbook.info Mohammed J. Zaki 1 Wagner Meira Jr. 2 1 Department of Computer Science Rensselaer Polytechnic Institute, Troy, NY, USA
More informationKernel-based Approximation. Methods using MATLAB. Gregory Fasshauer. Interdisciplinary Mathematical Sciences. Michael McCourt.
SINGAPORE SHANGHAI Vol TAIPEI - Interdisciplinary Mathematical Sciences 19 Kernel-based Approximation Methods using MATLAB Gregory Fasshauer Illinois Institute of Technology, USA Michael McCourt University
More informationData Mining Techniques
Data Mining Techniques CS 622 - Section 2 - Spring 27 Pre-final Review Jan-Willem van de Meent Feedback Feedback https://goo.gl/er7eo8 (also posted on Piazza) Also, please fill out your TRACE evaluations!
More informationA New Spectral Technique Using Normalized Adjacency Matrices for Graph Matching 1
CHAPTER-3 A New Spectral Technique Using Normalized Adjacency Matrices for Graph Matching Graph matching problem has found many applications in areas as diverse as chemical structure analysis, pattern
More informationContribution from: Springer Verlag Berlin Heidelberg 2005 ISBN
Contribution from: Mathematical Physics Studies Vol. 7 Perspectives in Analysis Essays in Honor of Lennart Carleson s 75th Birthday Michael Benedicks, Peter W. Jones, Stanislav Smirnov (Eds.) Springer
More informationSpectral Clustering. Zitao Liu
Spectral Clustering Zitao Liu Agenda Brief Clustering Review Similarity Graph Graph Laplacian Spectral Clustering Algorithm Graph Cut Point of View Random Walk Point of View Perturbation Theory Point of
More informationLecture 19: Isometries, Positive operators, Polar and singular value decompositions; Unitary matrices and classical groups; Previews (1)
Lecture 19: Isometries, Positive operators, Polar and singular value decompositions; Unitary matrices and classical groups; Previews (1) Travis Schedler Thurs, Nov 18, 2010 (version: Wed, Nov 17, 2:15
More informationLecture 1: Graphs, Adjacency Matrices, Graph Laplacian
Lecture 1: Graphs, Adjacency Matrices, Graph Laplacian Radu Balan January 31, 2017 G = (V, E) An undirected graph G is given by two pieces of information: a set of vertices V and a set of edges E, G =
More informationQualifying Examination Winter 2017
Qualifying Examination Winter 2017 Examination Committee: Anne Greenbaum, Hong Qian, Eric Shea-Brown Day 1, Tuesday, December 12, 9:30-12:30, LEW 208 You have three hours to complete this exam. Work all
More informationAPPENDIX A. Background Mathematics. A.1 Linear Algebra. Vector algebra. Let x denote the n-dimensional column vector with components x 1 x 2.
APPENDIX A Background Mathematics A. Linear Algebra A.. Vector algebra Let x denote the n-dimensional column vector with components 0 x x 2 B C @. A x n Definition 6 (scalar product). The scalar product
More informationSpectral Geometry of Riemann Surfaces
Spectral Geometry of Riemann Surfaces These are rough notes on Spectral Geometry and their application to hyperbolic riemann surfaces. They are based on Buser s text Geometry and Spectra of Compact Riemann
More informationData-dependent representations: Laplacian Eigenmaps
Data-dependent representations: Laplacian Eigenmaps November 4, 2015 Data Organization and Manifold Learning There are many techniques for Data Organization and Manifold Learning, e.g., Principal Component
More informationSpectral Generative Models for Graphs
Spectral Generative Models for Graphs David White and Richard C. Wilson Department of Computer Science University of York Heslington, York, UK wilson@cs.york.ac.uk Abstract Generative models are well known
More informationFundamentals of Matrices
Maschinelles Lernen II Fundamentals of Matrices Christoph Sawade/Niels Landwehr/Blaine Nelson Tobias Scheffer Matrix Examples Recap: Data Linear Model: f i x = w i T x Let X = x x n be the data matrix
More informationProperties of Matrices and Operations on Matrices
Properties of Matrices and Operations on Matrices A common data structure for statistical analysis is a rectangular array or matris. Rows represent individual observational units, or just observations,
More informationCHAPTER 5. THE KERNEL METHOD 138
CHAPTER 5. THE KERNEL METHOD 138 Chapter 5 The Kernel Method Before we can mine data, it is important to first find a suitable data representation that facilitates data analysis. For example, for complex
More informationSecond-Order Inference for Gaussian Random Curves
Second-Order Inference for Gaussian Random Curves With Application to DNA Minicircles Victor Panaretos David Kraus John Maddocks Ecole Polytechnique Fédérale de Lausanne Panaretos, Kraus, Maddocks (EPFL)
More informationAn Algorithmist s Toolkit September 10, Lecture 1
18.409 An Algorithmist s Toolkit September 10, 2009 Lecture 1 Lecturer: Jonathan Kelner Scribe: Jesse Geneson (2009) 1 Overview The class s goals, requirements, and policies were introduced, and topics
More informationUnsupervised Machine Learning and Data Mining. DS 5230 / DS Fall Lecture 7. Jan-Willem van de Meent
Unsupervised Machine Learning and Data Mining DS 5230 / DS 4420 - Fall 2018 Lecture 7 Jan-Willem van de Meent DIMENSIONALITY REDUCTION Borrowing from: Percy Liang (Stanford) Dimensionality Reduction Goal:
More informationLinear algebra and applications to graphs Part 1
Linear algebra and applications to graphs Part 1 Written up by Mikhail Belkin and Moon Duchin Instructor: Laszlo Babai June 17, 2001 1 Basic Linear Algebra Exercise 1.1 Let V and W be linear subspaces
More informationSpectral Theory of Unsigned and Signed Graphs Applications to Graph Clustering: a Survey
Spectral Theory of Unsigned and Signed Graphs Applications to Graph Clustering: a Survey Jean Gallier Department of Computer and Information Science University of Pennsylvania Philadelphia, PA 19104, USA
More informationGraph Matching & Information. Geometry. Towards a Quantum Approach. David Emms
Graph Matching & Information Geometry Towards a Quantum Approach David Emms Overview Graph matching problem. Quantum algorithms. Classical approaches. How these could help towards a Quantum approach. Graphs
More information5 Quiver Representations
5 Quiver Representations 5. Problems Problem 5.. Field embeddings. Recall that k(y,..., y m ) denotes the field of rational functions of y,..., y m over a field k. Let f : k[x,..., x n ] k(y,..., y m )
More informationKarhunen-Loeve Expansion and Optimal Low-Rank Model for Spatial Processes
TTU, October 26, 2012 p. 1/3 Karhunen-Loeve Expansion and Optimal Low-Rank Model for Spatial Processes Hao Zhang Department of Statistics Department of Forestry and Natural Resources Purdue University
More informationCOMS 4721: Machine Learning for Data Science Lecture 19, 4/6/2017
COMS 4721: Machine Learning for Data Science Lecture 19, 4/6/2017 Prof. John Paisley Department of Electrical Engineering & Data Science Institute Columbia University PRINCIPAL COMPONENT ANALYSIS DIMENSIONALITY
More informationConnection of Local Linear Embedding, ISOMAP, and Kernel Principal Component Analysis
Connection of Local Linear Embedding, ISOMAP, and Kernel Principal Component Analysis Alvina Goh Vision Reading Group 13 October 2005 Connection of Local Linear Embedding, ISOMAP, and Kernel Principal
More informationSVD, Power method, and Planted Graph problems (+ eigenvalues of random matrices)
Chapter 14 SVD, Power method, and Planted Graph problems (+ eigenvalues of random matrices) Today we continue the topic of low-dimensional approximation to datasets and matrices. Last time we saw the singular
More informationMachine Learning for Data Science (CS4786) Lecture 11
Machine Learning for Data Science (CS4786) Lecture 11 Spectral clustering Course Webpage : http://www.cs.cornell.edu/courses/cs4786/2016sp/ ANNOUNCEMENT 1 Assignment P1 the Diagnostic assignment 1 will
More informationIntroduction to Machine Learning
10-701 Introduction to Machine Learning PCA Slides based on 18-661 Fall 2018 PCA Raw data can be Complex, High-dimensional To understand a phenomenon we measure various related quantities If we knew what
More informationMachine Learning. Principal Components Analysis. Le Song. CSE6740/CS7641/ISYE6740, Fall 2012
Machine Learning CSE6740/CS7641/ISYE6740, Fall 2012 Principal Components Analysis Le Song Lecture 22, Nov 13, 2012 Based on slides from Eric Xing, CMU Reading: Chap 12.1, CB book 1 2 Factor or Component
More informationManifold Regularization
9.520: Statistical Learning Theory and Applications arch 3rd, 200 anifold Regularization Lecturer: Lorenzo Rosasco Scribe: Hooyoung Chung Introduction In this lecture we introduce a class of learning algorithms,
More information