DOCSLIB.ORG

Inner Products and Norms (Part III)

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Inner Products and Norms (Part III) Inner Products and Norms (part III) Prof. Dan A. Simovici UMB 1 / 74 Outline 1 Approximating Subspaces 2 Gram Matrices 3 The Gram-Schmidt Orthogonalization Algorithm 4 QR Decomposition of Matrices 5 Gram-Schmidt Algorithm in R 2 / 74 Approximating Subspaces Definition A subspace T of a inner product linear space is an approximating subspace if for every x 2 L there is a unique element in T that is closest to x. Theorem Let T be a subspace in the inner product space L. If x 2 L and t 2 T , then x − t 2 T ? if and only if t is the unique element of T closest to x. 3 / 74 Approximating Subspaces Proof Suppose that x − t 2 T ?. Then, for any u 2 T we have k x − u k2=k (x − t) + (t − u) k2=k x − t k2 + k t − u k2; by observing that x − t 2 T ? and t − u 2 T and applying Pythagora's 2 2 Theorem to x − t and t − u. Therefore, we have k x − u k >k x − t k , so t is the unique element of T closest to x. 4 / 74 Approximating Subspaces Proof (cont'd) Conversely, suppose that t is the unique element of T closest to x and x − t 62 T ?, that is, there exists u 2 T such that (x − t; u) 6= 0. This implies, of course, that u 6= 0L. We have k x − (t + au) k2=k x − t − au k2=k x − t k2 −2(x − t; au) + jaj2 k u k2 : 2 2 Since k x − (t + au) k >k x − t k (by the definition of t), we have 2 2 1 −2(x − t; au) + jaj k u k > 0 for every a 2 F. For a = kuk2 (x − t; u) we have: 1 1 −2(x − t; (x − t; u)u) + j (x − t; u)j2 k u k2 k u k2 k u k2 1 1 = −2(x − t; (x − t; u)u) + j (x − t; u)j2 k u k2 k u k2 k u k2 j(x − t; u)j2 j(x − t; u)j2 = −2 + k u k2 k u k2 j(x − t; u)j2 = − 0; k u k2 > which is a contradiction. 5 / 74 Approximating Subspaces Theorem A subspace T of an inner product linear space L is an approximating ? subspace L if and only if L = T T . 6 / 74 Approximating Subspaces Proof Let T be an approximating subspace of L and let x 2 L. We have x − t 2 T ?, where t is the element of T that best approximates x. If y = x − t, we can write x uniquely as x = t + y, where t 2 T and ? ? y 2 T , so L = T T . ? Conversely, suppose that L = T T , where T is a subspace of L. Every x 2 L can be uniquely written as x = t + y, where t 2 T and y 2 T ?, so x − t 2 T ?. Thus t is the element in T that is closest to x, so T is an approximating subspace of L. 7 / 74 Approximating Subspaces Theorem Any subspace T of a finite-dimensional inner product linear space L is an approximating subspace of L. 8 / 74 Approximating Subspaces ? Let T be a subspace of L. It suffices to show that L = T T . ? If T = f0Lg, then T = L and the statement is immediate. Therefore, we can assume that T 6= f0Lg. We need to verify only that every x 2 L can be uniquely written as a sum x = t + v, where t 2 T and v 2 T ?. Let t1;:::; tm be an orthonormal basis of T , that is, a basis such that ( 1 if i = j; (ti ; tj ) = 0 otherwise, for 1 6 i; j > m. Define t = (x; t1)t1 + ··· + (x; tm)tm and v = x − t. 9 / 74 Approximating Subspaces Proof (cont'd) The vector v is orthogonal to every vector ti because (v; ti ) = (x − t; ti ) = (x; ti ) − (t; ti ) = 0: Therefore v 2 T ? and x has the necessary decomposition. To prove that the decomposition is unique suppose that x = s + w, where s 2 T and ? w 2 T?. Since s + w = t + v we have s − t = v − w 2 T \ T = f0Lg, which implies s = t and w = v. 10 / 74 Approximating Subspaces Theorem Let T be a subspace of an inner product space L. We have (T ?)? = T . 11 / 74 Approximating Subspaces Observe that T ⊆ (T ?)?. Indeed, if t 2 T , then (t; z) = 0 for every z 2 T ?, so t 2 (T ?)?. To prove the reverse inclusion, let x 2 (T ?)?. We can write x = u + v, where u 2 T and v 2 T ?, so x − u = v 2 T ?. Since T ⊆ (T ?)?, we have u 2 (T ?)?, so x − u 2 (T ?)?. Consequently, x − u 2 T ? \ (T ?)? = f0g, so x = u 2 T . Thus, (T ?)? ⊆ T , which concludes the argument. 12 / 74 Approximating Subspaces Corollary n ? ? Let Z be a subset of R . We have (Z ) = hZi. 13 / 74 Approximating Subspaces Proof n ? ? Let Z be a subset of R . Since Z ⊆ hZi it follows that hZi ⊆ Z . Let ? now y 2 Z and let z = a1z1 + ··· + apzp 2 hZi, where z1;:::; zp 2 Z. Since (y; z) = a1(y; z1) + ··· + ap(y; zp) = 0; it follows that y 2 hZi?. Thus, we have Z ? = hZi?. ? ? ? ? n This allows us to write (Z ) = (hZi ) . Since hZi is a subspace of R , we have (hZi?)? = hZi, so (Z ?)? = hZi. 14 / 74 Approximating Subspaces Let W = fw1;:::; wng be a basis in the real n-dimensional inner product space L. If x = x1w1 + ··· + xnwn and y = y1w1 + ··· + ynwn, then n n X X (x; y) = xi yj (wi ; wj ); i=1 j=1 due to the bilinearity of the inner product. n×n Let A = (aij ) 2 R be the matrix defined by aij = (wi ; wj ) for 1 6 i; j 6 n. The symmetry of the inner product implies that the matrix A itself is symmetric. Now, the inner product can be expressed as 0 1 y1 B . C (x; y) = (x1;:::; xn)A @ . A : yn We refer to A as the matrix associated with W . 15 / 74 Approximating Subspaces Theorem n Let S be a subspace of R such that dim(S) = k. There exists a matrix n×k A 2 R having orthonormal columns such that S = Ran(A). 16 / 74 Approximating Subspaces Proof Let v1;:::; vk be an orthonormal basis of S. Define the matrix A as A = (v1;:::; vk ). We have x 2 S, if and only if x = a1v1 + ··· + ak vk , which is equivalent to x = Aa. This amounts to x 2 Ran(A), so S = Ran(A). 17 / 74 Approximating Subspaces For an orthonormal basis in an n-dimensional space, the associated matrix is the diagonal matrix In. In this case, we have (x; y) = x1y1 + x2y2 + ··· + xnyn for x; y 2 L. Observe that if W = fw1;:::; wng is an orthonormal set and x 2 hWi, which means that x = a1w1 + ··· + anwn, then ai = (x; wi ) for 1 6 i 6 n. 18 / 74 Approximating Subspaces Let W = fw1;:::; wng be an orthonormal set and let x 2 hWi. The equality x = (x; w1)w1 + ··· + (x; wn)wn is the Fourier expansion of x with respect to the orthonormal set W . Furthermore, we have Parseval's equality: n 2 X 2 k x k = (x; x) = (x; wi ) : i=1 Thus, if 1 6 q 6 n we have q X 2 2 (x; wi ) 6k x k : i=1 19 / 74 Gram Matrices Gram Matrices Definition Let V = (v1;:::; vm) be a sequence of vectors in an inner product space. m×m The Gram matrix of this sequence is the matrix GV = (gij ) 2 R defined by gij = (vi ; vj ) for 1 6 i; j 6 m. Note that GV is a symmetric matrix. 20 / 74 Gram Matrices Theorem Let V = (v1;:::; vm) be a sequence of m vectors in an inner product linear space (L; (·; ·)). If fv1;:::; vmg is linearly independent, then the Gram matrix GV is positive definite. 21 / 74 Gram Matrices Proof m Suppose that V is linearly independent and let x 2 R . We have m m 0 X X x GVx = xi (vi ; vj )xj i=1 j=1 0 m m 1 X X = @ xi vi ; xj vj A i=1 j=1 m 2 X = xi vi > 0: i=1 0 Therefore, if x GVx = 0, we have x1v1 + ··· + xnvn = 0. Since fv1;:::; vmg is linearly independent it follows that x1 = ··· = xm = 0, so x = 0n. Thus, GV is indeed, positive definite. 22 / 74 Gram Matrices Example S Let S = fx1;:::; xng be a finite set, L a C-linear space and L be the linear space that consists of function defined on S with values in L. We defined the linear basis of this space as fe1;:::; eng consisting of the functions ( 1 if x = xi ; ei (x) = 0 otherwise; for 1 6 i 6 n. If E = (e1;:::; en), the Gram matrix of E is positive definite Pn Pn and the inner product of two functions f = i=1 ai ei and g = j=1 bj ej is n n ! X X (f ; g) = ai ei ; bj ej i=1 i=1 n n X X = ai (ei ; ej )bj : i=1 i=1 23 / 74 Gram Matrices The Gram matrix of an arbitrary sequence is positive semidefinite, as the reader can easily verify.
Load more

Recommended publications
  • Tight Frames and Their Symmetries
    Technical Report 9 December 2003 Tight Frames and their Symmetries Richard Vale, Shayne Waldron Department of Mathematics, University of Auckland, Private Bag 92019, Auckland, New Zealand e–mail: [email protected] (http:www.math.auckland.ac.nz/˜waldron) e–mail: [email protected] ABSTRACT The aim of this paper is to investigate symmetry properties of tight frames, with a view to constructing tight frames of orthogonal polynomials in several variables which share the symmetries of the weight function, and other similar applications. This is achieved by using representation theory to give methods for constructing tight frames as orbits of groups of unitary transformations acting on a given finite-dimensional Hilbert space. Along the way, we show that a tight frame is determined by its Gram matrix and discuss how the symmetries of a tight frame are related to its Gram matrix. We also give a complete classification of those tight frames which arise as orbits of an abelian group of symmetries. Key Words: Tight frames, isometric tight frames, Gram matrix, multivariate orthogonal polynomials, symmetry groups, harmonic frames, representation theory, wavelets AMS (MOS) Subject Classifications: primary 05B20, 33C50, 20C15, 42C15, sec- ondary 52B15, 42C40 0 1. Introduction u1 u2 u3 2 The three equally spaced unit vectors u1, u2, u3 in IR provide the following redundant representation 2 3 f = f, u u , f IR2, (1.1) 3 h ji j ∀ ∈ j=1 X which is the simplest example of a tight frame. Such representations arose in the study of nonharmonic Fourier series in L2(IR) (see Duffin and Schaeffer [DS52]) and have recently been used extensively in the theory of wavelets (see, e.g., Daubechies [D92]).
    [Show full text]
  • Week 8-9. Inner Product Spaces. (Revised Version) Section 3.1 Dot Product As an Inner Product
    Math 2051 W2008 Margo Kondratieva Week 8-9. Inner product spaces. (revised version) Section 3.1 Dot product as an inner product. Consider a linear (vector) space V . (Let us restrict ourselves to only real spaces that is we will not deal with complex numbers and vectors.) De¯nition 1. An inner product on V is a function which assigns a real number, denoted by < ~u;~v> to every pair of vectors ~u;~v 2 V such that (1) < ~u;~v>=< ~v; ~u> for all ~u;~v 2 V ; (2) < ~u + ~v; ~w>=< ~u;~w> + < ~v; ~w> for all ~u;~v; ~w 2 V ; (3) < k~u;~v>= k < ~u;~v> for any k 2 R and ~u;~v 2 V . (4) < ~v;~v>¸ 0 for all ~v 2 V , and < ~v;~v>= 0 only for ~v = ~0. De¯nition 2. Inner product space is a vector space equipped with an inner product. Pn It is straightforward to check that the dot product introduces by ~u ¢ ~v = j=1 ujvj is an inner product. You are advised to verify all the properties listed in the de¯nition, as an exercise. The dot product is also called Euclidian inner product. De¯nition 3. Euclidian vector space is Rn equipped with Euclidian inner product < ~u;~v>= ~u¢~v. De¯nition 4. A square matrix A is called positive de¯nite if ~vT A~v> 0 for any vector ~v 6= ~0. · ¸ 2 0 Problem 1. Show that is positive de¯nite. 0 3 Solution: Take ~v = (x; y)T . Then ~vT A~v = 2x2 + 3y2 > 0 for (x; y) 6= (0; 0).
    [Show full text]
  • Gram Matrix and Orthogonality in Frames 1
    U.P.B. Sci. Bull., Series A, Vol. 80, Iss. 1, 2018 ISSN 1223-7027 GRAM MATRIX AND ORTHOGONALITY IN FRAMES Abolhassan FEREYDOONI1 and Elnaz OSGOOEI 2 In this paper, we aim at introducing a criterion that determines if f figi2I is a Bessel sequence, a frame or a Riesz sequence or not any of these, based on the norms and the inner products of the elements in f figi2I. In the cases of Riesz and Bessel sequences, we introduced a criterion but in the case of a frame, we did not find any answers. This criterion will be shown by K(f figi2I). Using the criterion introduced, some interesting extensions of orthogonality will be presented. Keywords: Frames, Bessel sequences, Orthonormal basis, Riesz bases, Gram matrix MSC2010: Primary 42C15, 47A05. 1. Preliminaries Frames are generalizations of orthonormal bases, but, more than orthonormal bases, they have shown their ability and stability in the representation of functions [1, 4, 10, 11]. The frames have been deeply studied from an abstract point of view. The results of such studies have been used in concrete frames such as Gabor and Wavelet frames which are very important from a practical point of view [2, 9, 5, 8]. An orthonormal set feng in a Hilbert space is characterized by a simple relation hem;eni = dm;n: In the other words, the Gram matrix is the identity matrix. Moreover, feng is an orthonor- mal basis if spanfeng = H. But, for frames the situation is more complicated; i.e., the Gram Matrix has no such a simple form.
    [Show full text]
  • Uniqueness of Low-Rank Matrix Completion by Rigidity Theory
    UNIQUENESS OF LOW-RANK MATRIX COMPLETION BY RIGIDITY THEORY AMIT SINGER∗ AND MIHAI CUCURINGU† Abstract. The problem of completing a low-rank matrix from a subset of its entries is often encountered in the analysis of incomplete data sets exhibiting an underlying factor model with applications in collaborative filtering, computer vision and control. Most recent work had been focused on constructing efficient algorithms for exact or approximate recovery of the missing matrix entries and proving lower bounds for the number of known entries that guarantee a successful recovery with high probability. A related problem from both the mathematical and algorithmic point of view is the distance geometry problem of realizing points in a Euclidean space from a given subset of their pairwise distances. Rigidity theory answers basic questions regarding the uniqueness of the realization satisfying a given partial set of distances. We observe that basic ideas and tools of rigidity theory can be adapted to determine uniqueness of low-rank matrix completion, where inner products play the role that distances play in rigidity theory. This observation leads to efficient randomized algorithms for testing necessary and sufficient conditions for local completion and for testing sufficient conditions for global completion. Crucial to our analysis is a new matrix, which we call the completion matrix, that serves as the analogue of the rigidity matrix. Key words. Low rank matrices, missing values, rigidity theory, iterative methods, collaborative filtering. AMS subject classifications. 05C10, 05C75, 15A48 1. Introduction. Can the missing entries of an incomplete real valued matrix be recovered? Clearly, a matrix can be completed in an infinite number of ways by replacing the missing entries with arbitrary values.
    [Show full text]
  • QUADRATIC FORMS and DEFINITE MATRICES 1.1. Definition of A
    QUADRATIC FORMS AND DEFINITE MATRICES 1. DEFINITION AND CLASSIFICATION OF QUADRATIC FORMS 1.1. Definition of a quadratic form. Let A denote an n x n symmetric matrix with real entries and let x denote an n x 1 column vector. Then Q = x’Ax is said to be a quadratic form. Note that a11 ··· a1n . x1 Q = x´Ax =(x1...xn) . xn an1 ··· ann P a1ixi . =(x1,x2, ··· ,xn) . P anixi 2 (1) = a11x1 + a12x1x2 + ... + a1nx1xn 2 + a21x2x1 + a22x2 + ... + a2nx2xn + ... + ... + ... 2 + an1xnx1 + an2xnx2 + ... + annxn = Pi ≤ j aij xi xj For example, consider the matrix 12 A = 21 and the vector x. Q is given by 0 12x1 Q = x Ax =[x1 x2] 21 x2 x1 =[x1 +2x2 2 x1 + x2 ] x2 2 2 = x1 +2x1 x2 +2x1 x2 + x2 2 2 = x1 +4x1 x2 + x2 1.2. Classification of the quadratic form Q = x0Ax: A quadratic form is said to be: a: negative definite: Q<0 when x =06 b: negative semidefinite: Q ≤ 0 for all x and Q =0for some x =06 c: positive definite: Q>0 when x =06 d: positive semidefinite: Q ≥ 0 for all x and Q = 0 for some x =06 e: indefinite: Q>0 for some x and Q<0 for some other x Date: September 14, 2004. 1 2 QUADRATIC FORMS AND DEFINITE MATRICES Consider as an example the 3x3 diagonal matrix D below and a general 3 element vector x. 100 D = 020 004 The general quadratic form is given by 100 x1 0 Q = x Ax =[x1 x2 x3] 020 x2 004 x3 x1 =[x 2 x 4 x ] x2 1 2 3 x3 2 2 2 = x1 +2x2 +4x3 Note that for any real vector x =06 , that Q will be positive, because the square of any number is positive, the coefficients of the squared terms are positive and the sum of positive numbers is always positive.
    [Show full text]
  • 8.3 Positive Definite Matrices
    8.3. Positive Definite Matrices 433 Exercise 8.2.25 Show that every 2 2 orthog- [Hint: If a2 + b2 = 1, then a = cos θ and b = sinθ for × cos θ sinθ some angle θ.] onal matrix has the form − or sinθ cosθ cos θ sin θ Exercise 8.2.26 Use Theorem 8.2.5 to show that every for some angle θ. sinθ cosθ symmetric matrix is orthogonally diagonalizable. − 8.3 Positive Definite Matrices All the eigenvalues of any symmetric matrix are real; this section is about the case in which the eigenvalues are positive. These matrices, which arise whenever optimization (maximum and minimum) problems are encountered, have countless applications throughout science and engineering. They also arise in statistics (for example, in factor analysis used in the social sciences) and in geometry (see Section 8.9). We will encounter them again in Chapter 10 when describing all inner products in Rn. Definition 8.5 Positive Definite Matrices A square matrix is called positive definite if it is symmetric and all its eigenvalues λ are positive, that is λ > 0. Because these matrices are symmetric, the principal axes theorem plays a central role in the theory. Theorem 8.3.1 If A is positive definite, then it is invertible and det A > 0. Proof. If A is n n and the eigenvalues are λ1, λ2, ..., λn, then det A = λ1λ2 λn > 0 by the principal axes theorem (or× the corollary to Theorem 8.2.5). ··· If x is a column in Rn and A is any real n n matrix, we view the 1 1 matrix xT Ax as a real number.
    [Show full text]
  • Exceptional Collections in Surface-Like Categories 11
    EXCEPTIONAL COLLECTIONS IN SURFACE-LIKE CATEGORIES ALEXANDER KUZNETSOV Abstract. We provide a categorical framework for recent results of Markus Perling on combinatorics of exceptional collections on numerically rational surfaces. Using it we simplify and generalize some of Perling’s results as well as Vial’s criterion for existence of a numerical exceptional collection. 1. Introduction A beautiful recent paper of Markus Perling [Pe] proves that any numerically exceptional collection of maximal length in the derived category of a numerically rational surface (i.e., a surface with zero irreg- ularity and geometric genus) can be transformed by mutations into an exceptional collection consisting of objects of rank 1. In this note we provide a categorical framework that allows to extend and simplify Perling’s result. For this we introduce a notion of a surface-like category. Roughly speaking, it is a triangulated cate- T num T gory whose numerical Grothendieck group K0 ( ), considered as an abelian group with a bilinear form (Euler form), behaves similarly to the numerical Grothendieck group of a smooth projective sur- face, see Definition 3.1. Of course, for any smooth projective surface X its derived category D(X) is surface-like. However, there are surface-like categories of different nature, for instance derived categories of noncommutative surfaces and of even-dimensional Calabi–Yau varieties turn out to be surface-like (Example 3.4). Also, some subcategories of surface-like categories are surface-like. Thus, the notion of a surface-like category is indeed more general. In fact, all results of this paper have a numerical nature, so instead of considering categories, we pass directly to their numerical Grothendieck groups.
    [Show full text]
  • Applications of the Wronskian and Gram Matrices of {Fie”K?
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Elsevier - Publisher Connector Applications of the Wronskian and Gram Matrices of {fie”k? Robert E. Hartwig Mathematics Department North Carolina State University Raleigh, North Carolina 27650 Submitted by S. Barn&t ABSTRACT A review is made of some of the fundamental properties of the sequence of functions {t’b’}, k=l,..., s, i=O ,..., m,_,, with distinct X,. In particular it is shown how the Wronskian and Gram matrices of this sequence of functions appear naturally in such fields as spectral matrix theory, controllability, and Lyapunov stability theory. 1. INTRODUCTION Let #(A) = II;,,< A-hk)“‘k be a complex polynomial with distinct roots x 1,.. ., A,, and let m=ml+ . +m,. It is well known [9] that the functions {tieXk’}, k=l,..., s, i=O,l,..., mkpl, form a fundamental (i.e., linearly inde- pendent) solution set to the differential equation m)Y(t)=O> (1) where D = $(.). It is the purpose of this review to illustrate some of the important properties of this independent solution set. In particular we shall show that both the Wronskian matrix and the Gram matrix play a dominant role in certain applications of matrix theory, such as spectral theory, controllability, and the Lyapunov stability theory. Few of the results in this paper are new; however, the proofs we give are novel and shed some new light on how the various concepts are interrelated, and on why certain results work the way they do. LINEAR ALGEBRA ANDITSAPPLlCATIONS43:229-241(1982) 229 C Elsevier Science Publishing Co., Inc., 1982 52 Vanderbilt Ave., New York, NY 10017 0024.3795/82/020229 + 13$02.50 230 ROBERT E.
    [Show full text]
  • On the Nyström Method for Approximating a Gram Matrix For
    Journal of Machine Learning Research 6 (2005) 2153–2175 Submitted 4/05; Published 12/05 On the Nystrom¨ Method for Approximating a Gram Matrix for Improved Kernel-Based Learning Petros Drineas [email protected] Department of Computer Science Rensselaer Polytechnic Institute Troy, NY 12180, USA Michael W. Mahoney [email protected] Department of Mathematics Yale University New Haven, CT 06520, USA Editor: Nello Cristianini Abstract A problem for many kernel-based methods is that the amount of computation required to find the solution scales as O(n3), where n is the number of training examples. We develop and analyze an algorithm to compute an easily-interpretable low-rank approximation to an n n Gram matrix G such that computations of interest may be performed more rapidly. The approximation× is of ˜ + T the form Gk = CWk C , where C is a matrix consisting of a small number c of columns of G and Wk is the best rank-k approximation to W, the matrix formed by the intersection between those c columns of G and the corresponding c rows of G. An important aspect of the algorithm is the probability distribution used to randomly sample the columns; we will use a judiciously-chosen and data-dependent nonuniform probability distribution. Let and denote the spectral norm k·k2 k·kF and the Frobenius norm, respectively, of a matrix, and let Gk be the best rank-k approximation to G. We prove that by choosing O(k/ε4) columns n + T ε 2 G CWk C ξ G Gk ξ + ∑ Gii, − ≤ k − k i=1 both in expectation and with high probability, for both ξ = 2, F, and for all k : 0 k rank(W).
    [Show full text]
  • Local Law for Random Gram Matrices
    Local law for random Gram matrices Johannes Alt∗ László Erdős∗ Torben Krüger∗ IST Austria IST Austria IST Austria [email protected] [email protected] [email protected] We prove a local law in the bulk of the spectrum for random Gram matrices XX∗, a generalization of sample covariance matrices, where X is a large matrix with independent, centered entries with arbitrary variances. The limiting eigenvalue density that generalizes the Marchenko-Pastur law is determined by solving a system of nonlinear equations. Our entrywise and averaged local laws are on the optimal scale with the optimal error bounds. They hold both in the square case (hard edge) and in the properly rectangular case (soft edge). In the latter case we also establish a macroscopic gap away from zero in the spectrum of XX∗. Keywords: Capacity of MIMO channels, Marchenko-Pastur law, hard edge, soft edge, general variance profile AMS Subject Classification: 60B20, 15B52 1. Introduction Random matrices were introduced in pioneering works by Wishart [43] and Wigner [42] for applications in mathematical statistics and nuclear physics, respectively. Wigner argued that the energy level statistics of large atomic nuclei could be described by the eigenvalues of a large Wigner matrix, i.e., a hermitian matrix N H = (hij )i,j=1 with centered, identically distributed and independent entries (up to the symmetry constraint H = H∗). He proved that the empirical spectral measure (or density of states) converges to the semicircle law as the dimension of the matrix N goes to infinity. Moreover, he postulated that the statistics of the gaps between consecutive eigenvalues depend only on the symmetry type of the matrix and are independent of the distribution of the entries in the large N limit.
    [Show full text]
  • 7 Spectral Properties of Matrices
    7 Spectral Properties of Matrices 7.1 Introduction The existence of directions that are preserved by linear transformations (which are referred to as eigenvectors) has been discovered by L. Euler in his study of movements of rigid bodies. This work was continued by Lagrange, Cauchy, Fourier, and Hermite. The study of eigenvectors and eigenvalues acquired in- creasing significance through its applications in heat propagation and stability theory. Later, Hilbert initiated the study of eigenvalue in functional analysis (in the theory of integral operators). He introduced the terms of eigenvalue and eigenvector. The term eigenvalue is a German-English hybrid formed from the German word eigen which means “own” and the English word “value”. It is interesting that Cauchy referred to the same concept as characteristic value and the term characteristic polynomial of a matrix (which we introduce in Definition 7.1) was derived from this naming. We present the notions of geometric and algebraic multiplicities of eigen- values, examine properties of spectra of special matrices, discuss variational characterizations of spectra and the relationships between matrix norms and eigenvalues. We conclude this chapter with a section dedicated to singular values of matrices. 7.2 Eigenvalues and Eigenvectors Let A Cn×n be a square matrix. An eigenpair of A is a pair (λ, x) C (Cn∈ 0 ) such that Ax = λx. We refer to λ is an eigenvalue and to ∈x is× an eigenvector−{ } . The set of eigenvalues of A is the spectrum of A and will be denoted by spec(A). If (λ, x) is an eigenpair of A, the linear system Ax = λx has a non-trivial solution in x.
    [Show full text]
  • Amath/Math 516 First Homework Set Solutions
    AMATH/MATH 516 FIRST HOMEWORK SET SOLUTIONS The purpose of this problem set is to have you brush up and further develop your multi-variable calculus and linear algebra skills. The problem set will be very difficult for some and straightforward for others. If you are having any difficulty, please feel free to discuss the problems with me at any time. Don't delay in starting work on these problems! 1. Let Q be an n × n symmetric positive definite matrix. The following fact for symmetric matrices can be used to answer the questions in this problem. Fact: If M is a real symmetric n×n matrix, then there is a real orthogonal n×n matrix U (U T U = I) T and a real diagonal matrix Λ = diag(λ1; λ2; : : : ; λn) such that M = UΛU . (a) Show that the eigenvalues of Q2 are the square of the eigenvalues of Q. Note that λ is an eigenvalue of Q if and only if there is some vector v such that Qv = λv. Then Q2v = Qλv = λ2v, so λ2 is an eigenvalue of Q2. (b) If λ1 ≥ λ2 ≥ · · · ≥ λn are the eigen values of Q, show that 2 T 2 n λnkuk2 ≤ u Qu ≤ λ1kuk2 8 u 2 IR : T T Pn 2 We have u Qu = u UΛUu = i=1 λikuk2. The result follows immediately from bounds on the λi. (c) If 0 < λ < λ¯ are such that 2 T ¯ 2 n λkuk2 ≤ u Qu ≤ λkuk2 8 u 2 IR ; then all of the eigenvalues of Q must lie in the interval [λ; λ¯].
    [Show full text]