DOCSLIB.ORG

NOTE on MATH 4010: FUNCTIONAL ANALYSIS Throughout This Note, All

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
NOTE on MATH 4010: FUNCTIONAL ANALYSIS Throughout This Note, All NOTE ON MATH 4010: FUNCTIONAL ANALYSIS CHI-WAI LEUNG Throughout this note, all spaces X; Y; :: are normed spaces over the field K = R or C. Let BX := fx 2 X : kxk ≤ 1g and SX := fx 2 X : kxk = 1g denote the closed unit ball and the unit sphere of X respectively. We write a sequence of numbers as a function x : f1; 2; :::g ! K. 1. Classical Normed spaces The following examples are important classes in the study of functional analysis. Example 1.1. Put 1 c0 := f(x(i)) : x(i) 2 K; lim jx(i)j = 0g and ` := f(x(i)) : x(i) 2 K; sup x(i) < 1g: i 1 1 Then c0 is a subspace of ` . The sup-norm k · k1 on ` is defined by kxk1 := supi jx(i)j for 1 1 1 x 2 ` . Then ` is a Banach space and (c0; k · k1) is a closed subspace of ` (Check !) and hence c0 is also a Banach space too. Let c00 := f(x(i)) : there are only finitly many x(i)'s are non-zerog: Also, c00 is endowed with the sup-norm defined above. Then c00 is not a Banach space (Why?) but it is dense in c0, that is, c00 = c0 (Check!). Example 1.2. For 1 ≤ p < 1. Put 1 p X p ` := f(x(i)) : x(i) 2 K; jx(i)j < 1g: i=1 1 1 p X p p p Also, ` is equipped with the norm kxkp := ( jx(i)j ) p for x 2 ` . Then ` becomes a Banach i=1 space under the norm k · kp. Example 1.3. Let X be a locally compact Hausdorff space, for example, K. Let C0(X) be the space of all continuous K-valued functions f on X which are vanish at infinity, that is, for every " > 0, there is a compact subset D of X such that jf(x)j < " for all x 2 X n D. Now C0(X) is endowed with the sup-norm, that is, kfk1 = sup jf(x)j x2X for every f 2 C0(X). Then C0(X) is a Banach space. (Try to prove this fact for the case X = R. Just use the knowledge from MATH 2060 !!!) 2. Finite Dimensional Normed Spaces We say that two norms k · k and k · k0 on a vector space X are equivalent, write k · k ∼ k · k0, if 0 there are positive numbers c1 and c2 such that c1k · k ≤ k · k ≤ c2k · k on X. Date: February 11, 2017. 1 2 CHI-WAI LEUNG 1 Example 2.1. Consider the norms k · k1 and k · k1 on ` . We are going to show that k · k1 and k · k1 are not equivalent. In fact, if we put xn(i) := (1; 1=2; :::; 1=n; 0; 0; ::::) for n; i = 1; 2:::. Then 1 xn 2 ` for all n. Notice that (xn) is a Cauchy sequence with respect to the norm k · k1 but it is 1 not a Cauchy sequence with respect to the norm k · k1. Hence k · k1 k · k1 on ` . Proposition 2.2. All norms on a finite dimensional vector space are equivalent. Proof. Let X be a finite dimensional vector space and let fe1; :::; eng be a vector base of X. For Pn Pn each x = i=1 αiei for αi 2 K, define kxk0 = i=1 jαij. Then k · k0 is a norm X. The result is obtained by showing that all norms k · k on X are equivalent to k · k0. Pn Notice that for each x = αiei 2 X, we have kxk ≤ ( max keik)kxk0. It remains to find i=1 1≤i≤n n c > 0 such that ck · k0 ≤ k · k. In fact, let K be equipped with the sup-norm k · k1, that is n k(α1; :::; αn)k1 = max1≤1≤n jαij. Define a real-valued function f on the unit sphere SKn of K by f :(α1; :::; αn) 2 SKn 7! kα1e1 + ··· + αnenk: Notice that the map f is continuous and f > 0. It is clear that SKn is compact with respect to the n sup-norm k · k1 on K . Hence, there is c > 0 such that f(α) ≥ c > 0 for all α 2 SKn . This gives kxk ≥ ckxk0 for all x 2 X as desired. The proof is finished. Corollary 2.3. We have the following assertions. (i) All finite dimensional normed spaces are Banach spaces. Consequently, any finite dimen- sional subspace of a normed space must be closed. (ii) The closed unit ball of any finite dimensional normed space is compact. Proof. Let (X; k · k) be a finite dimensional normed space. With the notation as in the proof of Proposition 2.2 above, we see that k · k must be equivalent to the norm k · k0. It is clear that X is complete with respect to the norm k · k0 and so is complete in the original norm k · k. The Part (i) follows. For Part (ii), it is clear that the compactness of the closed unit ball of X is equivalent to saying that any closed and bounded subset being compact. Therefore, Part (ii) follows from the simple observation that any closed and bounded subset of X with respect to the norm k · k0 is compact. The proof is complete. In the rest of this section, we are going to show the converse of Corollary 2.3(ii) also holds. Before this result, we need the following useful result. Lemma 2.4. Riesz's Lemma: Let Y be a closed proper subspace of a normed space X. Then for each θ 2 (0; 1), there is an element x0 2 SX such that d(x0;Y ) := inffkx − yk : y 2 Y g ≥ θ. Proof. Let u 2 X − Y and d := inffku − yk : y 2 Y g. Notice that since Y is closed, d > 0 d and hence, we have 0 < d < θ because 0 < θ < 1. This implies that there is y0 2 Y such that d u−y0 0 < d ≤ ku − y0k < . Now put x0 := 2 S . We are going to show that x0 is as desired. θ ku−y0k X Indeed, let y 2 Y . Since y0 + ku − y0ky 2 Y , we have 1 kx0 − yk = ku − (y0 + ku − y0ky)k ≥ d=ku − y0k > θ: ku − y0k So, d(x0;Y ) ≥ θ. Remark 2.5. The Riesz's lemma does not hold when θ = 1. The following example can be found in the Diestel's interesting book without proof (see [2, Chapter 1 Ex.3(i)]). R 1 Let X = fx 2 C([0; 1]; R): x(0) = 0g and Y = fy 2 X : 0 y(t)dt = 0g. Both X and Y are endowed with the sup-norm. Notice that Y is a closed proper subspace of X. We are going to show 3 that for any x 2 SX , there is y 2 Y such that kx − yk1 < 1. Thus, the Riesz's Lemma does not hold as θ = 1 in this case. In fact, let x 2 SX . Since x(0) = 0 with kxk1 = 1, we can find 0 < a < 1=4 such that jx(t)j ≤ 1=4 for all t 2 [0; a]. Notice that since x is uniform continuous on [a; 1], for any 0 < " < 1=4, there 0 0 is δ > 0 such that jx(t) − x(t )j < "=4 when jt − t j < δ. Now we find a partition a = t0 < t1 < ··· < tn = 1 with tk − tk−1 < δ for all k = 1; 2; :::n and jx(tk)j < 1 for all k = 1; 2:::; n − 1. Then 0 0 supfjx(t) − x(t )j : t; t 2 [tk−1; tk]g < "=4. We let pk−1 := supft 2 [tk−1; tk]: xj[tk−1;t] > −1 + "g if it exists, otherwise, put pk−1 := tk−1. Similarly, let qk := infft 2 [tk−1; tk]: xj[t;tk] > −1 + "g if it exists, otherwise, put qk := tk. So, one can find a continuous function φ on [a; 1] such that 8 " if t 2 [t ; t ] and xj > −1 + ": > k−1 k [tk−1;tk] > <−" if t 2 [pk−1; qk] and xj[tk−1;tk] ≯ −1 + ": φ(t) = −2" (t − tk−1) + " if xj[t ;t ] ≯ −1 + " and tk−1 < t < pk−1: > pk−1−tk−1 k−1 k > 2" :> (t − tk) + " if xj −1 + " and qk < t < tk: tk−qk [tk−1;tk] ≯ Notice that if xj[tk−1;tk] ≯ −1 + ", then tk−1 < pk−1 or qk < tk. So, kxj[a;1] − φk1 < 1. R 1 It is because kφk1 < 2", we have j a φ(t)dtj ≤ 2"(1 − a). On the other hand, as jx(t)j < 1=4 on [0; a], so if we further choose " small enough such that (1 − a)(2") < a=4, then we can find a continuous function y1 on [0; a] such that jy1(t)j < 1=4 on [0; a] with ; y1(0) = 0; y1(a) = x(a) and R a R 1 0 y1(t)dt = − a φ(t)dt. Now we define y = y1 on [0; a] and y = φ on [a; 1]. Then ky − xk1 < 1 and y 2 Y is as desired. Theorem 2.6. X is a finite dimensional normed space if and only if the closed unit ball BX of X is compact. Proof. The necessary condition has been shown by Proposition 2.3(ii). Now assume that X is of infinite dimension. Fix an element x1 2 SX .
Load more

Recommended publications
  • Spheres in Infinite-Dimensional Normed Spaces Are Lipschitz Contractible
    PROCEEDINGS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 88. Number 3, July 1983 SPHERES IN INFINITE-DIMENSIONAL NORMED SPACES ARE LIPSCHITZ CONTRACTIBLE Y. BENYAMINI1 AND Y. STERNFELD Abstract. Let X be an infinite-dimensional normed space. We prove the following: (i) The unit sphere {x G X: || x II = 1} is Lipschitz contractible. (ii) There is a Lipschitz retraction from the unit ball of JConto the unit sphere. (iii) There is a Lipschitz map T of the unit ball into itself without an approximate fixed point, i.e. inffjjc - Tx\\: \\x\\ « 1} > 0. Introduction. Let A be a normed space, and let Bx — {jc G X: \\x\\ < 1} and Sx = {jc G X: || jc || = 1} be its unit ball and unit sphere, respectively. Brouwer's fixed point theorem states that when X is finite dimensional, every continuous self-map of Bx admits a fixed point. Two equivalent formulations of this theorem are the following. 1. There is no continuous retraction from Bx onto Sx. 2. Sx is not contractible, i.e., the identity map on Sx is not homotopic to a constant map. It is well known that none of these three theorems hold in infinite-dimensional spaces (see e.g. [1]). The natural generalization to infinite-dimensional spaces, however, would seem to require the maps to be uniformly-continuous and not merely continuous. Indeed in the finite-dimensional case this condition is automatically satisfied. In this article we show that the above three theorems fail, in the infinite-dimen- sional case, even under the strongest uniform-continuity condition, namely, for maps satisfying a Lipschitz condition.
    [Show full text]
  • Examples of Manifolds
    Examples of Manifolds Example 1 (Open Subset of IRn) Any open subset, O, of IRn is a manifold of dimension n. One possible atlas is A = (O, ϕid) , where ϕid is the identity map. That is, ϕid(x) = x. n Of course one possible choice of O is IR itself. Example 2 (The Circle) The circle S1 = (x,y) ∈ IR2 x2 + y2 = 1 is a manifold of dimension one. One possible atlas is A = {(U , ϕ ), (U , ϕ )} where 1 1 1 2 1 y U1 = S \{(−1, 0)} ϕ1(x,y) = arctan x with − π < ϕ1(x,y) <π ϕ1 1 y U2 = S \{(1, 0)} ϕ2(x,y) = arctan x with 0 < ϕ2(x,y) < 2π U1 n n n+1 2 2 Example 3 (S ) The n–sphere S = x =(x1, ··· ,xn+1) ∈ IR x1 +···+xn+1 =1 n A U , ϕ , V ,ψ i n is a manifold of dimension . One possible atlas is 1 = ( i i) ( i i) 1 ≤ ≤ +1 where, for each 1 ≤ i ≤ n + 1, n Ui = (x1, ··· ,xn+1) ∈ S xi > 0 ϕi(x1, ··· ,xn+1)=(x1, ··· ,xi−1,xi+1, ··· ,xn+1) n Vi = (x1, ··· ,xn+1) ∈ S xi < 0 ψi(x1, ··· ,xn+1)=(x1, ··· ,xi−1,xi+1, ··· ,xn+1) n So both ϕi and ψi project onto IR , viewed as the hyperplane xi = 0. Another possible atlas is n n A2 = S \{(0, ··· , 0, 1)}, ϕ , S \{(0, ··· , 0, −1)},ψ where 2x1 2xn ϕ(x , ··· ,xn ) = , ··· , 1 +1 1−xn+1 1−xn+1 2x1 2xn ψ(x , ··· ,xn ) = , ··· , 1 +1 1+xn+1 1+xn+1 are the stereographic projections from the north and south poles, respectively.
    [Show full text]
  • Higher Dimensional Conundra
    Higher Dimensional Conundra Steven G. Krantz1 Abstract: In recent years, especially in the subject of harmonic analysis, there has been interest in geometric phenomena of RN as N → +∞. In the present paper we examine several spe- cific geometric phenomena in Euclidean space and calculate the asymptotics as the dimension gets large. 0 Introduction Typically when we do geometry we concentrate on a specific venue in a particular space. Often the context is Euclidean space, and often the work is done in R2 or R3. But in modern work there are many aspects of analysis that are linked to concrete aspects of geometry. And there is often interest in rendering the ideas in Hilbert space or some other infinite dimensional setting. Thus we want to see how the finite-dimensional result in RN changes as N → +∞. In the present paper we study some particular aspects of the geometry of RN and their asymptotic behavior as N →∞. We choose these particular examples because the results are surprising or especially interesting. We may hope that they will lead to further studies. It is a pleasure to thank Richard W. Cottle for a careful reading of an early draft of this paper and for useful comments. 1 Volume in RN Let us begin by calculating the volume of the unit ball in RN and the surface area of its bounding unit sphere. We let ΩN denote the former and ωN−1 denote the latter. In addition, we let Γ(x) be the celebrated Gamma function of L. Euler. It is a helpful intuition (which is literally true when x is an 1We are happy to thank the American Institute of Mathematics for its hospitality and support during this work.
    [Show full text]
  • MTH 304: General Topology Semester 2, 2017-2018
    MTH 304: General Topology Semester 2, 2017-2018 Dr. Prahlad Vaidyanathan Contents I. Continuous Functions3 1. First Definitions................................3 2. Open Sets...................................4 3. Continuity by Open Sets...........................6 II. Topological Spaces8 1. Definition and Examples...........................8 2. Metric Spaces................................. 11 3. Basis for a topology.............................. 16 4. The Product Topology on X × Y ...................... 18 Q 5. The Product Topology on Xα ....................... 20 6. Closed Sets.................................. 22 7. Continuous Functions............................. 27 8. The Quotient Topology............................ 30 III.Properties of Topological Spaces 36 1. The Hausdorff property............................ 36 2. Connectedness................................. 37 3. Path Connectedness............................. 41 4. Local Connectedness............................. 44 5. Compactness................................. 46 6. Compact Subsets of Rn ............................ 50 7. Continuous Functions on Compact Sets................... 52 8. Compactness in Metric Spaces........................ 56 9. Local Compactness.............................. 59 IV.Separation Axioms 62 1. Regular Spaces................................ 62 2. Normal Spaces................................ 64 3. Tietze's extension Theorem......................... 67 4. Urysohn Metrization Theorem........................ 71 5. Imbedding of Manifolds..........................
    [Show full text]
  • Convolution on the N-Sphere with Application to PDF Modeling Ivan Dokmanic´, Student Member, IEEE, and Davor Petrinovic´, Member, IEEE
    IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 58, NO. 3, MARCH 2010 1157 Convolution on the n-Sphere With Application to PDF Modeling Ivan Dokmanic´, Student Member, IEEE, and Davor Petrinovic´, Member, IEEE Abstract—In this paper, we derive an explicit form of the convo- emphasis on wavelet transform in [8]–[12]. Computation of the lution theorem for functions on an -sphere. Our motivation comes Fourier transform and convolution on groups is studied within from the design of a probability density estimator for -dimen- the theory of noncommutative harmonic analysis. Examples sional random vectors. We propose a probability density function (pdf) estimation method that uses the derived convolution result of applications of noncommutative harmonic analysis in engi- on . Random samples are mapped onto the -sphere and esti- neering are analysis of the motion of a rigid body, workspace mation is performed in the new domain by convolving the samples generation in robotics, template matching in image processing, with the smoothing kernel density. The convolution is carried out tomography, etc. A comprehensive list with accompanying in the spectral domain. Samples are mapped between the -sphere theory and explanations is given in [13]. and the -dimensional Euclidean space by the generalized stereo- graphic projection. We apply the proposed model to several syn- Statistics of random vectors whose realizations are observed thetic and real-world data sets and discuss the results. along manifolds embedded in Euclidean spaces are commonly termed directional statistics. An excellent review may be found Index Terms—Convolution, density estimation, hypersphere, hy- perspherical harmonics, -sphere, rotations, spherical harmonics. in [14]. It is of interest to develop tools for the directional sta- tistics in analogy with the ordinary Euclidean.
    [Show full text]
  • Minkowski Products of Unit Quaternion Sets 1 Introduction
    Minkowski products of unit quaternion sets 1 Introduction The Minkowski sum A⊕B of two point sets A; B 2 Rn is the set of all points generated [16] by the vector sums of points chosen independently from those sets, i.e., A ⊕ B := f a + b : a 2 A and b 2 B g : (1) The Minkowski sum has applications in computer graphics, geometric design, image processing, and related fields [9, 11, 12, 13, 14, 15, 20]. The validity of the definition (1) in Rn for all n ≥ 1 stems from the straightforward extension of the vector sum a + b to higher{dimensional Euclidean spaces. However, to define a Minkowski product set A ⊗ B := f a b : a 2 A and b 2 B g ; (2) it is necessary to specify products of points in Rn. In the case n = 1, this is simply the real{number product | the resulting algebra of point sets in R1 is called interval arithmetic [17, 18] and is used to monitor the propagation of uncertainty through computations in which the initial operands (and possibly also the arithmetic operations) are not precisely determined. A natural realization of the Minkowski product (2) in R2 may be achieved [7] by interpreting the points a and b as complex numbers, with a b being the usual complex{number product. Algorithms to compute Minkowski products of complex{number sets have been formulated [6], and extended to determine Minkowski roots and powers [3, 8] of complex sets; to evaluate polynomials specified by complex{set coefficients and arguments [4]; and to solve simple equations expressed in terms of complex{set coefficients and unknowns [5].
    [Show full text]
  • General Topology
    General Topology Tom Leinster 2014{15 Contents A Topological spaces2 A1 Review of metric spaces.......................2 A2 The definition of topological space.................8 A3 Metrics versus topologies....................... 13 A4 Continuous maps........................... 17 A5 When are two spaces homeomorphic?................ 22 A6 Topological properties........................ 26 A7 Bases................................. 28 A8 Closure and interior......................... 31 A9 Subspaces (new spaces from old, 1)................. 35 A10 Products (new spaces from old, 2)................. 39 A11 Quotients (new spaces from old, 3)................. 43 A12 Review of ChapterA......................... 48 B Compactness 51 B1 The definition of compactness.................... 51 B2 Closed bounded intervals are compact............... 55 B3 Compactness and subspaces..................... 56 B4 Compactness and products..................... 58 B5 The compact subsets of Rn ..................... 59 B6 Compactness and quotients (and images)............. 61 B7 Compact metric spaces........................ 64 C Connectedness 68 C1 The definition of connectedness................... 68 C2 Connected subsets of the real line.................. 72 C3 Path-connectedness.......................... 76 C4 Connected-components and path-components........... 80 1 Chapter A Topological spaces A1 Review of metric spaces For the lecture of Thursday, 18 September 2014 Almost everything in this section should have been covered in Honours Analysis, with the possible exception of some of the examples. For that reason, this lecture is longer than usual. Definition A1.1 Let X be a set. A metric on X is a function d: X × X ! [0; 1) with the following three properties: • d(x; y) = 0 () x = y, for x; y 2 X; • d(x; y) + d(y; z) ≥ d(x; z) for all x; y; z 2 X (triangle inequality); • d(x; y) = d(y; x) for all x; y 2 X (symmetry).
    [Show full text]
  • Ball, Deborah Loewenberg Developing Mathematics
    DOCUMENT RESUME ED 399 262 SP 036 932 AUTHOR Ball, Deborah Loewenberg TITLE Developing Mathematics Reform: What Don't We Know about Teacher Learning--But Would Make Good Working Hypotheses? NCRTL Craft Paper 95-4. INSTITUTION National Center for Research on Teacher Learning, East Lansing, MI. SPONS AGENCY Office of Educational Research and Improvement (ED), Washington, DC. PUB DATE Oct 95 NOTE 53p.; Paper presented at a conference on Teacher Enhancement in Mathematics K-6 (Arlington, VA, November 18-20, 1994). AVAILABLE FROMNational Center for Research on Teacher Learning, 116 Erickson Hall, Michigan State University, East Lansing, MI 48824-1034. PUB TYPE Reports Research/Technical (143) Speeches /Conference Papers (150) EDRS PRICE MF01/PC03 Plus Postage. DESCRIPTORS *Educational Change; *Educational Policy; Elementary School Mathematics; Elementary Secondary Education; *Faculty Development; *Knowledge Base for Teaching; *Mathematics Instruction; Mathematics Teachers; Secondary School Mathematics; *Teacher Improvement ABSTRACT This paper examines what teacher educators, policymakers, and teachers think they know about the current mathematics reforms and what it takes to help teachers engage with these reforms. The analysis is organized around three issues:(1) the "it" envisioned by the reforms;(2) what teachers (and others) bring to learning "it";(3) what is known and believed about teacher learning. The second part of the paper deals with what is not known about the reforms and how to help a larger number of teachers engage productively with the reforms. This section of the paper confronts the current pressures to "scale up" reform efforts to "reach" more teachers. Arguing that what is known is not sufficient to meet the demand for scaling up, three potential hypotheses about teacher learning and professional development are proposed that might serve to meet the demand to offer more teachers opportunities for learning.
    [Show full text]
  • GEOMETRY Contents 1. Euclidean Geometry 2 1.1. Metric Spaces 2 1.2
    GEOMETRY JOUNI PARKKONEN Contents 1. Euclidean geometry 2 1.1. Metric spaces 2 1.2. Euclidean space 2 1.3. Isometries 4 2. The sphere 7 2.1. More on cosine and sine laws 10 2.2. Isometries 11 2.3. Classification of isometries 12 3. Map projections 14 3.1. The latitude-longitude map 14 3.2. Stereographic projection 14 3.3. Inversion 14 3.4. Mercator’s projection 16 3.5. Some Riemannian geometry. 18 3.6. Cylindrical projection 18 4. Triangles in the sphere 19 5. Minkowski space 21 5.1. Bilinear forms and Minkowski space 21 5.2. The orthogonal group 22 6. Hyperbolic space 24 6.1. Isometries 25 7. Models of hyperbolic space 30 7.1. Klein’s model 30 7.2. Poincaré’s ball model 30 7.3. The upper halfspace model 31 8. Some geometry and techniques 32 8.1. Triangles 32 8.2. Geodesic lines and isometries 33 8.3. Balls 35 9. Riemannian metrics, area and volume 36 Last update: December 12, 2014. 1 1. Euclidean 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.
    [Show full text]
  • Lp Unit Spheres and the Α-Geometries: Questions and Perspectives
    entropy Article Lp Unit Spheres and the a-Geometries: Questions and Perspectives Paolo Gibilisco Department of Economics and Finance, University of Rome “Tor Vergata”, Via Columbia 2, 00133 Rome, Italy; [email protected] Received: 12 November 2020; Accepted: 10 December 2020; Published: 14 December 2020 Abstract: In Information Geometry, the unit sphere of Lp spaces plays an important role. In this paper, the aim is list a number of open problems, in classical and quantum IG, which are related to Lp geometry. Keywords: Lp spheres; a-geometries; a-Proudman–Johnson equations Gentlemen: there’s lots of room left in Lp spaces. 1. Introduction Chentsov theorem is the fundamental theorem in Information Geometry. After Rao’s remark on the geometric nature of the Fisher Information (in what follows shortly FI), it is Chentsov who showed that on the simplex of the probability vectors, up to scalars, FI is the unique Riemannian geometry, which “contract under noise” (to have an idea of recent developments about this see [1]). So FI appears as the “natural” Riemannian geometry over the manifolds of density vectors, namely over 1 n Pn := fr 2 R j ∑ ri = 1, ri > 0g i Since FI is the pull-back of the map p r ! 2 r it is natural to study the geometries induced on the simplex of probability vectors by the embeddings 8 1 <p · r p p 2 [1, +¥) Ap(r) = :log(r) p = +¥ Setting 2 p = a 2 [−1, 1] 1 − a we call the corresponding geometries on the simplex of probability vectors a-geometries (first studied by Chentsov himself).
    [Show full text]
  • Simplices in the Euclidean Ball Matthieu Fradelizi, Grigoris Paouris, Carsten Schütt
    Simplices in the Euclidean ball Matthieu Fradelizi, Grigoris Paouris, Carsten Schütt To cite this version: Matthieu Fradelizi, Grigoris Paouris, Carsten Schütt. Simplices in the Euclidean ball. Canadian Mathematical Bulletin, 2011, 55 (3), pp.498-508. 10.4153/CMB-2011-142-1. hal-00731269 HAL Id: hal-00731269 https://hal-upec-upem.archives-ouvertes.fr/hal-00731269 Submitted on 12 Sep 2012 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Simplices in the Euclidean ball Matthieu Fradelizi, Grigoris Paouris,∗ Carsten Sch¨utt To appear in Canad. Math. Bull. Abstract We establish some inequalities for the second moment 1 2 |x|2dx |K| ZK of a convex body K under various assumptions on the position of K. 1 Introduction The starting point of this paper is the article [2], where it was shown that if all the extreme points of a convex body K in Rn have Euclidean norm greater than r > 0, then 1 r2 x 2dx > (1) K | |2 9n | | ZK where x 2 stands for the Euclidean norm of x and K for the volume of K. | | | | r2 We improve here this inequality showing that the optimal constant is , n + 2 with equality for the regular simplex, with vertices on the Euclidean sphere of radius r.
    [Show full text]
  • 17 Measure Concentration for the Sphere
    17 Measure Concentration for the Sphere In today’s lecture, we will prove the measure concentration theorem for the sphere. Recall that this was one of the vital steps in the analysis of the Arora-Rao-Vazirani approximation algorithm for sparsest cut. Most of the material in today’s lecture is adapted from Matousek’s book [Mat02, chapter 14] and Keith Ball’s lecture notes on convex geometry [Bal97]. n n−1 Notation: We will use the notation Bn to denote the ball of unit radius in R and S n to denote the sphere of unit radius in R . Let µ denote the normalized measure on the unit sphere (i.e., for any measurable set S ⊆ Sn−1, µ(A) denotes the ratio of the surface area of µ to the entire surface area of the sphere Sn−1). Recall that the n-dimensional volume of a ball n n n of radius r in R is given by the formula Vol(Bn) · r = vn · r where πn/2 vn = n Γ 2 + 1 Z ∞ where Γ(x) = tx−1e−tdt 0 n−1 The surface area of the unit sphere S is nvn. Theorem 17.1 (Measure Concentration for the Sphere Sn−1) Let A ⊆ Sn−1 be a mea- n−1 surable subset of the unit sphere S such that µ(A) = 1/2. Let Aδ denote the δ-neighborhood n−1 n−1 of A in S . i.e., Aδ = {x ∈ S |∃z ∈ A, ||x − z||2 ≤ δ}. Then, −nδ2/2 µ(Aδ) ≥ 1 − 2e . Thus, the above theorem states that if A is any set of measure 0.5, taking a step of even √ O (1/ n) around A covers almost 99% of the entire sphere.
    [Show full text]