DOCSLIB.ORG

Isometry and Automorphisms of Constant Dimension Codes

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Isometry and Automorphisms of Constant Dimension Codes Isometry and Automorphisms of Constant Dimension Codes Isometry and Automorphisms of Constant Dimension Codes Anna-Lena Trautmann Institute of Mathematics University of Zurich \Crypto and Coding" Z¨urich, March 12th 2012 1 / 25 Isometry and Automorphisms of Constant Dimension Codes Introduction 1 Introduction 2 Isometry of Random Network Codes 3 Isometry and Automorphisms of Known Code Constructions Spread codes Lifted rank-metric codes Orbit codes 2 / 25 isometry classes are equivalence classes automorphism groups of linear codes are useful for decoding automorphism groups are canonical representative of orbit codes Isometry and Automorphisms of Constant Dimension Codes Introduction Motivation constant dimension codes are used for random network coding 3 / 25 automorphism groups of linear codes are useful for decoding automorphism groups are canonical representative of orbit codes Isometry and Automorphisms of Constant Dimension Codes Introduction Motivation constant dimension codes are used for random network coding isometry classes are equivalence classes 3 / 25 automorphism groups are canonical representative of orbit codes Isometry and Automorphisms of Constant Dimension Codes Introduction Motivation constant dimension codes are used for random network coding isometry classes are equivalence classes automorphism groups of linear codes are useful for decoding 3 / 25 Isometry and Automorphisms of Constant Dimension Codes Introduction Motivation constant dimension codes are used for random network coding isometry classes are equivalence classes automorphism groups of linear codes are useful for decoding automorphism groups are canonical representative of orbit codes 3 / 25 Definition Subspace metric: dS(U; V) = dim(U + V) − dim(U\V) Injection metric: dI (U; V) = max(dim U; dim V) − dim(U\V) Isometry and Automorphisms of Constant Dimension Codes Introduction Random Network Codes Definition n n The projective geometry P(Fq ) is the set of all subspaces of Fq . n A random network code is a subset of P(Fq ). 4 / 25 Isometry and Automorphisms of Constant Dimension Codes Introduction Random Network Codes Definition n n The projective geometry P(Fq ) is the set of all subspaces of Fq . n A random network code is a subset of P(Fq ). Definition Subspace metric: dS(U; V) = dim(U + V) − dim(U\V) Injection metric: dI (U; V) = max(dim U; dim V) − dim(U\V) 4 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes 1 Introduction 2 Isometry of Random Network Codes 3 Isometry and Automorphisms of Known Code Constructions Spread codes Lifted rank-metric codes Orbit codes 5 / 25 Any isometry ι is injective: U 6= V () d(U; V) 6= 0 () d(ι(U); ι(V)) 6= 0 () ι(U) 6= ι(V) and hence, if the domain is equal to the codomain, bijective. The inverse map ι−1 is an isometry as well. Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Definition n n A distance-preserving map ι : P(Fq ) !P(Fq ) i.e. fulfilling n d(U; V) = d(ι(U); ι(V)) 8 U; V 2 P(Fq ): n is called an isometry on P(Fq ). 6 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Definition n n A distance-preserving map ι : P(Fq ) !P(Fq ) i.e. fulfilling n d(U; V) = d(ι(U); ι(V)) 8 U; V 2 P(Fq ): n is called an isometry on P(Fq ). Any isometry ι is injective: U 6= V () d(U; V) 6= 0 () d(ι(U); ι(V)) 6= 0 () ι(U) 6= ι(V) and hence, if the domain is equal to the codomain, bijective. The inverse map ι−1 is an isometry as well. 6 / 25 Lemma n Let ι be as before and U 2 P(Fq ) arbitrary. Then ι(f0g) = f0g =) dim(U) = d(f0g; U) = d(f0g; ι(U)) = dim(ι(U)) and on the other hand n n ι(f0g) = Fq =) dim(U) = d(f0g; U) = d(Fq ; ι(U)) = n−dim(ι(U)): The isometries with ι(f0g) = f0g are exactly the isometries that keep the dimension of a codeword. Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Lemma n n n If ι : P(Fq ) !P(Fq ) is an isometry, then ι(f0g) 2 f0g; Fq . 7 / 25 The isometries with ι(f0g) = f0g are exactly the isometries that keep the dimension of a codeword. Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Lemma n n n If ι : P(Fq ) !P(Fq ) is an isometry, then ι(f0g) 2 f0g; Fq . Lemma n Let ι be as before and U 2 P(Fq ) arbitrary. Then ι(f0g) = f0g =) dim(U) = d(f0g; U) = d(f0g; ι(U)) = dim(ι(U)) and on the other hand n n ι(f0g) = Fq =) dim(U) = d(f0g; U) = d(Fq ; ι(U)) = n−dim(ι(U)): 7 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Lemma n n n If ι : P(Fq ) !P(Fq ) is an isometry, then ι(f0g) 2 f0g; Fq . Lemma n Let ι be as before and U 2 P(Fq ) arbitrary. Then ι(f0g) = f0g =) dim(U) = d(f0g; U) = d(f0g; ι(U)) = dim(ι(U)) and on the other hand n n ι(f0g) = Fq =) dim(U) = d(f0g; U) = d(Fq ; ι(U)) = n−dim(ι(U)): The isometries with ι(f0g) = f0g are exactly the isometries that keep the dimension of a codeword. 7 / 25 Theorem n n For n > 2 a map ι : P(Fq ) !P(Fq ) is an order-preserving n bijection (with respect to the subset relation) of P(Fq ) if and only if it is an isometry with ι(f0g) = f0g. Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Theorem (Fundamental Theorem of Projective Geometry) n n Every order-preserving bijection f : P(Fq ) !P(Fq ), where n > 2, is induced by a semilinear transformation (A; α) 2 PΓLn = (GLn=Zn) o Aut(Fq) ∗ where Zn = fµIn j µ 2 Fqg is the set of scalar transformations. 8 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Theorem (Fundamental Theorem of Projective Geometry) n n Every order-preserving bijection f : P(Fq ) !P(Fq ), where n > 2, is induced by a semilinear transformation (A; α) 2 PΓLn = (GLn=Zn) o Aut(Fq) ∗ where Zn = fµIn j µ 2 Fqg is the set of scalar transformations. Theorem n n For n > 2 a map ι : P(Fq ) !P(Fq ) is an order-preserving n bijection (with respect to the subset relation) of P(Fq ) if and only if it is an isometry with ι(f0g) = f0g. 8 / 25 Remark: 1 All isometric codes are equivalent from a coding point of view (i.e. same rate and error correction capability). 2 There are more equivalence maps than order-preserving isometries. Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Definition n 1 Two codes C1; C2 ⊆ P(Fq ) are linearly isometric if there exists A 2 PGLn (or GLn) such that C1 = C2A. 2 We call C1 and C2 semilinearly isometric if there exists (A; α) 2 PΓLn (or ΓLn) such that C1 = C2(A; α). 9 / 25 2 There are more equivalence maps than order-preserving isometries. Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Definition n 1 Two codes C1; C2 ⊆ P(Fq ) are linearly isometric if there exists A 2 PGLn (or GLn) such that C1 = C2A. 2 We call C1 and C2 semilinearly isometric if there exists (A; α) 2 PΓLn (or ΓLn) such that C1 = C2(A; α). Remark: 1 All isometric codes are equivalent from a coding point of view (i.e. same rate and error correction capability). 9 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry of Random Network Codes Definition n 1 Two codes C1; C2 ⊆ P(Fq ) are linearly isometric if there exists A 2 PGLn (or GLn) such that C1 = C2A. 2 We call C1 and C2 semilinearly isometric if there exists (A; α) 2 PΓLn (or ΓLn) such that C1 = C2(A; α). Remark: 1 All isometric codes are equivalent from a coding point of view (i.e. same rate and error correction capability). 2 There are more equivalence maps than order-preserving isometries. 9 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry and Automorphisms of Known Code Constructions 1 Introduction 2 Isometry of Random Network Codes 3 Isometry and Automorphisms of Known Code Constructions Spread codes Lifted rank-metric codes Orbit codes 10 / 25 Definition The Grassmannian Gq(k; n) is the set of all k-dimensional n subspaces of Fq .A constant dimension code is a subset of Gq(k; n). Isometry and Automorphisms of Constant Dimension Codes Isometry and Automorphisms of Known Code Constructions Definition n For a given code C ⊆ P(Fq ), Aut(C) = fA 2 GLnjCA = Cg is called the (linear) automorphism group of the code. 11 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry and Automorphisms of Known Code Constructions Definition n For a given code C ⊆ P(Fq ), Aut(C) = fA 2 GLnjCA = Cg is called the (linear) automorphism group of the code. Definition The Grassmannian Gq(k; n) is the set of all k-dimensional n subspaces of Fq .A constant dimension code is a subset of Gq(k; n). 11 / 25 Isometry and Automorphisms of Constant Dimension Codes Isometry and Automorphisms of Known Code Constructions Spread codes 1 Introduction 2 Isometry of Random Network Codes 3 Isometry and Automorphisms of Known Code Constructions Spread codes Lifted rank-metric codes Orbit codes 12 / 25 l Proof: Since there is only one spread of lines in Fqk , different n Desarguesian spreads of Fq can only arise from the different k isomorphisms between Fqk and Fq .
Load more

Recommended publications
  • On Isometries on Some Banach Spaces: Part I
    Introduction Isometries on some important Banach spaces Hermitian projections Generalized bicircular projections Generalized n-circular projections On isometries on some Banach spaces { something old, something new, something borrowed, something blue, Part I Dijana Iliˇsevi´c University of Zagreb, Croatia Recent Trends in Operator Theory and Applications Memphis, TN, USA, May 3{5, 2018 Recent work of D.I. has been fully supported by the Croatian Science Foundation under the project IP-2016-06-1046. Dijana Iliˇsevi´c On isometries on some Banach spaces Introduction Isometries on some important Banach spaces Hermitian projections Generalized bicircular projections Generalized n-circular projections Something old, something new, something borrowed, something blue is referred to the collection of items that helps to guarantee fertility and prosperity. Dijana Iliˇsevi´c On isometries on some Banach spaces Introduction Isometries on some important Banach spaces Hermitian projections Generalized bicircular projections Generalized n-circular projections Isometries Isometries are maps between metric spaces which preserve distance between elements. Definition Let (X ; j · j) and (Y; k · k) be two normed spaces over the same field. A linear map ': X!Y is called a linear isometry if k'(x)k = jxj; x 2 X : We shall be interested in surjective linear isometries on Banach spaces. One of the main problems is to give explicit description of isometries on a particular space. Dijana Iliˇsevi´c On isometries on some Banach spaces Introduction Isometries on some important Banach spaces Hermitian projections Generalized bicircular projections Generalized n-circular projections Isometries Isometries are maps between metric spaces which preserve distance between elements. Definition Let (X ; j · j) and (Y; k · k) be two normed spaces over the same field.
    [Show full text]
  • Computing the Complexity of the Relation of Isometry Between Separable Banach Spaces
    Computing the complexity of the relation of isometry between separable Banach spaces Julien Melleray Abstract We compute here the Borel complexity of the relation of isometry between separable Banach spaces, using results of Gao, Kechris [1] and Mayer-Wolf [4]. We show that this relation is Borel bireducible to the universal relation for Borel actions of Polish groups. 1 Introduction Over the past fteen years or so, the theory of complexity of Borel equivalence relations has been a very active eld of research; in this paper, we compute the complexity of a relation of geometric nature, the relation of (linear) isom- etry between separable Banach spaces. Before stating precisely our result, we begin by quickly recalling the basic facts and denitions that we need in the following of the article; we refer the reader to [3] for a thorough introduction to the concepts and methods of descriptive set theory. Before going on with the proof and denitions, it is worth pointing out that this article mostly consists in putting together various results which were already known, and deduce from them after an easy manipulation the complexity of the relation of isometry between separable Banach spaces. Since this problem has been open for a rather long time, it still seems worth it to explain how this can be done, and give pointers to the literature. Still, it seems to the author that this will be mostly of interest to people with a knowledge of descriptive set theory, so below we don't recall descriptive set-theoretic notions but explain quickly what Lipschitz-free Banach spaces are and how that theory works.
    [Show full text]
  • Isometries and the Plane
    Chapter 1 Isometries of the Plane \For geometry, you know, is the gate of science, and the gate is so low and small that one can only enter it as a little child. (W. K. Clifford) The focus of this first chapter is the 2-dimensional real plane R2, in which a point P can be described by its coordinates: 2 P 2 R ;P = (x; y); x 2 R; y 2 R: Alternatively, we can describe P as a complex number by writing P = (x; y) = x + iy 2 C: 2 The plane R comes with a usual distance. If P1 = (x1; y1);P2 = (x2; y2) 2 R2 are two points in the plane, then p 2 2 d(P1;P2) = (x2 − x1) + (y2 − y1) : Note that this is consistent withp the complex notation. For P = x + iy 2 C, p 2 2 recall that jP j = x + y = P P , thus for two complex points P1 = x1 + iy1;P2 = x2 + iy2 2 C, we have q d(P1;P2) = jP2 − P1j = (P2 − P1)(P2 − P1) p 2 2 = j(x2 − x1) + i(y2 − y1)j = (x2 − x1) + (y2 − y1) ; where ( ) denotes the complex conjugation, i.e. x + iy = x − iy. We are now interested in planar transformations (that is, maps from R2 to R2) that preserve distances. 1 2 CHAPTER 1. ISOMETRIES OF THE PLANE Points in the Plane • A point P in the plane is a pair of real numbers P=(x,y). d(0,P)2 = x2+y2. • A point P=(x,y) in the plane can be seen as a complex number x+iy.
    [Show full text]
  • Isometry Test for Coordinate Transformations Sudesh Kumar Arora
    Isometry test for coordinate transformations Sudesh Kumar Arora Isometry test for coordinate transformations Sudesh Kumar Arora∗y Date January 30, 2020 Abstract A simple scheme is developed to test whether a given coordinate transformation of the metric tensor is an isometry or not. The test can also be used to predict the time dilation caused by any type of motion for a given space-time metric. Keywords and phrases: general relativity; coordinate transformation; diffeomorphism; isometry; symmetries. 1 Introduction Isometry is any coordinate transformation which preserves the distances. In special relativity, Minkowski metric has Poincare as the group of isometries, and the corresponding coordinate transformations leave the physical properties of the metric like distance invariant. In general relativity any infinitesimal trans- formation along Killing vector fields is an isometry. These are infinitesimal transformations generated by one parameter x0µ = xµ + V µ (1) Here V (x) = V µ@µ can be thought of like a vector field with components given as dx0µ V µ(x) = (x) (2) d All coordinate transformations generated such can be checked for isometry by calculating Lie deriva- tive of the metric tensors along the vector field V . If the Lie derivative vanishes then the transformation is an isometry (1; 2) and vector field V is called Killing Vector field. Killing vector fields are used to study the symmetries of the space-time metrics and the associated conserved quantities. But there are some challenges in using Killing equations for checking isometry. Coordinate transformations which are either discrete (not generated by the one-parameter ) or not possible to define as integral curves of a vectors field cannot be checked for isometry using this scheme.
    [Show full text]
  • What Is Symmetry? What Is an Isometry?
    2. SYMMETRY IN GEOMETRY 2.1. Isometries What is Symmetry? You probably have an intuitive idea of what symme- try means and can recognise it in various guises. For example, you can hopefully see that the letters N, O and M are symmetric, while the letter R is not. But probably, you can’t define in a mathematically pre- cise way what it means to be symmetric — this is something we’re going to address. Symmetry occurs very often in nature, a particular example being in the human body, as demonstrated by Leonardo da Vinci’s drawing of the Vitruvian Man on the right.1 The symmetry in the picture arises since it looks essentially the same when we flip it over. A particularly important observation about the drawing is that the distance from the Vitruvian Man’s left index finger to his left elbow is the same as the distance from his right index finger to his right elbow. Similarly, the distance from the Vitruvian Man’s left knee to his left eye is the same as the distance from his right knee to his right eye. We could go on and on writing down statements like this, but the point I’m trying to get at is that our intuitive notion of sym- metry — at least in geometry — is somehow tied up with the notion of distance. What is an Isometry? The flip in the previous discussion was a particular function which took points in the picture to other points in the picture. In particular, it did this in such a way that two points which were a certain distance apart would get mapped to two points which were the same distance apart.
    [Show full text]
  • ISOMETRIES of Rn 1. Introduction an Isometry of R N Is A
    ISOMETRIES OF Rn KEITH CONRAD 1. Introduction An isometry of Rn is a function h: Rn ! Rn that preserves the distance between vectors: jjh(v) − h(w)jj = jjv − wjj n p 2 2 for all v and w in R , where jj(x1; : : : ; xn)jj = x1 + ··· + xn. Example 1.1. The identity transformation: id(v) = v for all v 2 Rn. Example 1.2. Negation: − id(v) = −v for all v 2 Rn. n Example 1.3. Translation: fixing u 2 R , let tu(v) = v + u. Easily jjtu(v) − tu(w)jj = jjv − wjj. Example 1.4. Rotations around points and reflections across lines in the plane are isome- tries of R2. Formulas for these isometries will be given in Example 3.3 and Section4. The effects of a translation, rotation (around the origin) and reflection across a line in R2 are pictured below on sample line segments. The composition of two isometries of Rn is an isometry. Is every isometry invertible? It is clear that the three kinds of isometries pictured above (translations, rotations, reflections) are each invertible (translate by the negative vector, rotate by the opposite angle, reflect a second time across the same line). In Section2, we show the close link between isometries and the dot product on Rn, which is more convenient to use than distances due to its algebraic properties. Section3 is about the matrices that act as isometries on on Rn, called orthogonal matrices. Section 4 describes the isometries of R and R2 geometrically. In AppendixA, we will look more closely at reflections in Rn.
    [Show full text]
  • The Generic Isometry and Measure Preserving Homeomorphism Are Conjugate to Their Powers
    THE GENERIC ISOMETRY AND MEASURE PRESERVING HOMEOMORPHISM ARE CONJUGATE TO THEIR POWERS CHRISTIAN ROSENDAL Abstract. It is known that there is a comeagre set of mutually conjugate measure preserving homeomorphisms of Cantor space equipped with the coin- flipping probability measure, i.e., Haar measure. We show that the generic measure preserving homeomorphism is moreover conjugate to all of its pow- ers. It follows that the generic measure preserving homeomorphism extends to an action of (Q; +) by measure preserving homeomorphisms, and, in fact, to an action of the locally compact ring A of finite ad`eles. Similarly, S. Solecki has proved that there is a comeagre set of mutually conjugate isometries of the rational Urysohn metric space. We prove that these are all conjugate with their powers and therefore also embed into Q-actions. In fact, we extend these actions to actions of A as in the case of measure preserving homeomorphisms. We also consider a notion of topological similarity in Polish groups and use this to give simplified proofs of the meagreness of conjugacy classes in the automorphism group of the standard probability space and in the isometry group of the Urysohn metric space. Contents 1. Introduction 1 2. Powers of generic measure preserving homeomorphisms 4 2.1. Free amalgams of measured Boolean algebras 4 2.2. Roots of measure preserving homeomorphisms 6 2.3. The ring of finite ad`eles 9 2.4. Actions of A by measure preserving homeomorphisms 12 3. Powers of generic isometries 15 3.1. Free amalgams of metric spaces 15 3.2. Roots of isometries 15 3.3.
    [Show full text]
  • MATH 304 Linear Algebra Lecture 37: Orthogonal Matrices (Continued)
    MATH 304 Linear Algebra Lecture 37: Orthogonal matrices (continued). Rigid motions. Rotations in space. Orthogonal matrices Definition. A square matrix A is called orthogonal T T T 1 if AA = A A = I , i.e., A = A− . Theorem 1 If A is an n n orthogonal matrix, then × (i) columns of A form an orthonormal basis for Rn; (ii) rows of A also form an orthonormal basis for Rn. Idea of the proof: Entries of matrix ATA are dot products of columns of A. Entries of AAT are dot products of rows of A. Theorem 2 If A is an n n orthogonal matrix, × then (i) A is diagonalizable in the complexified vector space Cn; (ii) all eigenvalues λ of A satisfy λ = 1. | | cos φ sin φ Example. Aφ = − , φ R. sin φ cos φ ∈ AφAψ = Aφ+ψ • 1 T Aφ− = A φ = Aφ • − Aφ is orthogonal • iφ Eigenvalues: λ1 = cos φ + i sin φ = e , • iφ λ2 = cos φ i sin φ = e− . − Associated eigenvectors: v1 = (1, i), • − v2 = (1, i). λ2 = λ1 and v2 = v1. • 1 1 Vectors v1 and v2 form an orthonormal • √2 √2 basis for C2. Consider a linear operator L : Rn Rn, L(x)= Ax, → where A is an n n matrix. × Theorem The following conditions are equivalent: (i) L(x) = x for all x Rn; k k k k ∈ (ii) L(x) L(y)= x y for all x, y Rn; · · ∈ (iii) the transformation L preserves distance between points: L(x) L(y) = x y for all x, y Rn; k − k k − k ∈ (iv) L preserves length of vectors and angle between vectors; (v) the matrix A is orthogonal; (vi) the matrix of L relative to any orthonormal basis is orthogonal; (vii) L maps some orthonormal basis for Rn to another orthonormal basis; (viii) L maps any orthonormal basis for Rn to another orthonormal basis.
    [Show full text]
  • Contractions, Isometries and Some Properties of Inner-Product Spaces
    MATHEMATICS CONTRACTIONS, ISOMETRIES AND SOME PROPERTIES OF INNER-PRODUCT SPACES BY M. EDELSTEIN AND A. C. THOMPSON (Communicated by Prof. H. D. KLOOSTERMAN at the meeting of January 28, 1967) 1. INTRODUCTION In this paper we are concerned, among other things, with conditions under which contractions and isometries defined on certain subsets of a normed space are extendable to the whole space as the same type of mapping. (By a contraction and isometry we understand a mapping under which distances are not increased or are preserved respectively.) Our main interest is in possibly weak hypotheses which force the space to be an inner-product one. In Section 3 the notion of an isometric sequence is introduced. The definition of this concept is in terms of an isometry and, in real Hilbert space, a specific construction is exhibited extending this isometry to the whole space. Using this extension several properties of isometric sequences are described. 2. EXTENSION OF CONTRACTIONS AND ISOMETRIES Let E and F be normed spaces and ~ a nonempty family of nonempty subsets of E. The pair (E, F) is said to have the contraction (isometry) extension property with respect to ~. if for every DE~ and any con­ traction (isometry) T: D -+ F there is a contraction (isometry) 1': E -+ F which is an extension ofT. When~. in the above definition, is the family of all nonempty subsets of E, (E, F) is simply said to have the contraction (isometry) extension property. This last notion is due to S. 0. ScHoNBECK [9]. Answering a question of R. A. HIRSCHFELD [4] he also stated that whenever E and F are Banach spaces, F strictly convex, and (E, F) has the contraction extension property then both E and F are Hilbert spaces.
    [Show full text]
  • GROUPS and GEOMETRY 1. Euclidean and Spherical Geometry
    GROUPS AND GEOMETRY JOUNI PARKKONEN 1. Euclidean and spherical geometry 1.1. Metric spaces. A function d: X × X ! [0; +1[ is a metric in the nonempty set X if it satisfies the following properties (1) d(x; x) = 0 for all x 2 X and d(x; y) > 0 if x 6= y, (2) d(x; y) = d(y; x) for all x; y 2 X, and (3) d(x; y) ≤ d(x; z) + d(z; y) for all x; y; z 2 X (the triangle inequality). The pair (X; d) is a metric space. Open and closed balls in a metric space, continuity of maps between metric spaces and other “metric properties” are defined in the same way as in Euclidean space, using the metrics of X and Y instead of the Euclidean metric. If (X1; d1) and (X2; d2) are metric spaces, then a map i: X ! Y is an isometric embedding, if d2(i(x); i(y)) = d1(x; y) for all x; y 2 X1. If the isometric embedding i is a bijection, then it is called an isometry between X and Y . An isometry i: X ! X is called an isometry of X. The isometries of a metric space X form a group Isom(X), the isometry group of X, with the composition of mappings as the group law. A map i: X ! Y is a locally isometric embedding if each point x 2 X has a neighbourhood U such that the restriction of i to U is an isometric embedding. A (locally) isometric embedding i: I ! X is (1)a (locally) geodesic segment, if I ⊂ R is a (closed) bounded interval, (2)a (locally) geodesic ray, if I = [0; +1[, and (3)a (locally) geodesic line, if I = R.
    [Show full text]
  • Nearly Isometric Embedding by Relaxation
    Nearly Isometric Embedding by Relaxation James McQueen Marina Meila˘ Dominique Perrault-Joncas Department of Statistics Department of Statistics Google University of Washington University of Washington Seattle, WA 98103 Seattle, WA 98195 Seattle, WA 98195 [email protected] [email protected] [email protected] Abstract Many manifold learning algorithms aim to create embeddings with low or no dis- tortion (isometric). If the data has intrinsic dimension d, it is often impossible to obtain an isometric embedding in d dimensions, but possible in s>d dimensions. Yet, most geometry preserving algorithms cannot do the latter. This paper pro- poses an embedding algorithm to overcome this. The algorithm accepts as input, besides the dimension d, an embedding dimension s ≥ d. For any data embedding Y, we compute a Loss(Y), based on the push-forward Riemannian metric associ- ated with Y, which measures deviation of Y from from isometry. Riemannian Relaxation iteratively updates Y in order to decrease Loss(Y). The experiments confirm the superiority of our algorithm in obtaining low distortion embeddings. 1 Introduction, background and problem formulation Suppose we observe data points sampled from a smooth manifold M with intrinsic dimension d which is itself a submanifold of D-dimensional Euclidean space M ⊂ RD. The task of manifold learning is to provide a mapping φ : M →N (where N ⊂ Rs) of the manifold into lower dimensional space s ≪ D. According to the Whitney Embedding Theorem [11] we know that M can be embedded smoothly into R2d using one homeomorphism φ. Hence we seek one smooth map φ : M → Rs with d ≤ s ≤ 2d ≪ D.
    [Show full text]
  • On Isometry Groups of Pseudo-Riemannian Compact Lie Groups 3
    ON ISOMETRY GROUPS OF PSEUDO-RIEMANNIAN COMPACT LIE GROUPS FUHAI ZHU, ZHIQI CHEN, AND KE LIANG Abstract. Let G be a connected, simply-connected, compact simple Lie group. In this paper, we show that the isometry group of G with a left-invariant pseudo-Riemannan metric is compact. Furthermore, the identity component of the isometry group is compact if G is not simply-connected. 1. Introduction The study on isometry groups plays a fundamental role in the study of geometry, and there are beautiful results on isometry groups of Riemannian manifolds and Lorentzian manifolds, see [1, 2, 3, 6] and the references therein. But for a general pseudo-Riemannian manifold M, we know much less than that for Riemannian and Lorentzian cases. Let I (M) be the isometry group of M and Io(M) the identity component of I (M). An important result is given by Gromov in [4] that I (M) contains a closed connected normal abelian subgroup A such that I (M)/A is compact if M is compact, simply-connected and real analytic. Moreover, D’Ambra ([1]) proves that the isometry group I (M) of a compact, real analytic, simply connected Lorentzian manifold is compact, and gives an example of a simply connected compact pseudo-Riemannian manifold of type (7, 2) with a noncompact isometry group. In this paper, we focus on group manifolds, and prove the compactness of the isometry group of a connected, compact simple Lie group with a left-invariant pseudo-Riemannian metric. The paper is organized as follows. In section 2, we will give some notations and list some important results on isometry groups which are useful tools to prove the following theorems in section 3.
    [Show full text]