Applications of Central Limit Theorems over asymptotically measurable sets: Regression Models

Consider a regression model X_n = (ξ_{n, Yn}), n ∈ ℤ^d and assume that X_n is centered, φ is regular, ξ and Y are independent random fields, and that the Central Limit Theorem (CLT) holds for φ(ξ, y), for each y fixed, but that Y satisfies only the Strong Law of Large Numbers applied to (Y, Y_−n), for any n ∈ ℤ^d. Then the CLT holds for X. The basic technique is the application of the CLT over asymptotically measurable sets.     

Intersection patterns of convex sets

LetK ₁,…K_n be convex sets inR ^d. For 0≦i denote byf _ithe number of subsetsS of {1,2,…,n} of cardinalityi+1 that satisfy ∩{K _i∶i∈S}≠Ø. We prove:Theorem.If f _d+r=0 for somer r>=0, then {fx161-1} This inequality was conjectured by Katchalski and Perles. Equality holds, e.g., ifK ₁=…=K_r=R^d andK _r+1,…,K_n aren?r hyperplanes in general position inR ^d. The proof uses multilinear techniques (exterior algebra). Applications to convexity and to extremal set theory are given.     

Holomorphic Linearization of Commuting Germs of Holomorphic Maps

Let f ₁,…,f _h be h≥2 germs of biholomorphisms of ? ⁿ fixing the origin. We investigate the shape that a (formal) simultaneous linearization of the given germs can have, and we prove that if f ₁,…,f _h commute and their linear parts are almost simultaneously Jordanizable, then they are simultaneously formally linearizable. We next introduce a simultaneous Brjuno-type condition and prove that, in case the linear terms of the germs are diagonalizable, if the germs commute and our Brjuno-type condition holds, then they are holomorphically simultaneously linearizable. This answers a multi-dimensional version of a problem raised by Moser.     

The codimension of degenerate pencils

Let d_n[d_n(r)] denote the codimension of the set of pairs of n×n Hermitian [really symmetric] matrices (A, B) for which det(λI?A?xB)=p(λ,x) is a reducible polynomial. We prove that d_n(r)?n?1, d_n?_n?1 (n odd), d_n?n (n even). We conjecture that the equality holds in all three inequalities. We prove this conjecture for n=2,3.     

On the minimum length of ternary linear codes

From the geometrical point of view, we prove that [g ₃(6, d) + 1, 6, d]₃ codes exist for d = 118–123, 283–297 and that [g ₃(6, d), 6, d]₃ codes for d = 100, 341, 342 and [g ₃(6, d) + 1, 6, d]₃ codes for d = 130, 131, 132 do not exist, where ${g_3(k,\,d)=\sum_{i=0}^{k-1}\left\lceil d/3^i \right\rceil}$ . These determine the exact value of n ₃(6, d) for d = 100, 118–123, 130, 131, 132, 283–297, 341, 342, where n _q (k, d) is the minimum length n for which an [n, k, d] _q code exists.     

Necessary and sufficient consistency conditions for a recursive kernel regression estimate

A recursive kernel estimate ∑_{i = 1}ⁿ Y_iK⧸(x − X_i)h_i)⧸∑_{j = 1}ⁿ K((x − X_j)⧸h_j) of a regression m(x) = E{Y|X = x} calculated from independent observations (X₁, Y₁),…, (X_n, Y_n) of a pair (X, Y) of random variables is examined. ForE|Y|^{1 + δ} < ∞, δ > 0, the estimate is weakly pointwise consistent for almost all (μ) x ∈ R^d, μ is the probability measure of X, if and only if∑_i−1ⁿ h_i^d I_{hi > ɛ } ⧸ ∑_{j = 1}ⁿ h_j^d → 0 as n → ∞, all ɛ > 0, and∑_{i = 1}^∞ h_i^d = ∞, d is the dimension of X. For E|Y|^{1 + δ} < ∞, δ > 0, the estimate is strongly pointwise consistent for almost all (μ) x ∈ R^d, if and only if the same conditions hold. ForE|Y|^{1 + δ} < ∞, δ > 0, weak and strong consistency are equivalent. Similar results are given for complete convergence.     

Fast and accurate tensor approximation of a multivariate convolution with linear scaling in dimension

In the present paper we present the tensor-product approximation of a multidimensional convolution transform discretized via a collocation-projection scheme on uniform or composite refined grids. Examples of convolving kernels are provided by the classical Newton, Slater (exponential) and Yukawa potentials, 1/‖x‖, and with x∈R^d. For piecewise constant elements on the uniform grid of size n^d, we prove quadratic convergence O(h²) in the mesh parameter h=1/n, and then justify the Richardson extrapolation method on a sequence of grids that improves the order of approximation up to O(h³). A fast algorithm of complexity O(dR₁R₂nlogn) is described for tensor-product convolution on uniform/composite grids of size n^d, where R₁,R₂ are tensor ranks of convolving functions. We also present the tensor-product convolution scheme in the two-level Tucker canonical format and discuss the consequent rank reduction strategy. Finally, we give numerical illustrations confirming: (a) the approximation theory for convolution schemes of order O(h²) and O(h³); (b) linear-logarithmic scaling of 1D discrete convolution on composite grids; (c) linear-logarithmic scaling in n of our tensor-product convolution method on an n×n×n grid in the range n≤16384.     

A central limit theorem for a weighted power variation of a Gaussian process*

Let ρ be a real-valued function on [0, T], and let LSI(ρ) be a class of Gaussian processes over time interval [0, T], which need not have stationary increments but their incremental variance σ(s, t) is close to the values ρ(|t ? s|) as t → s uniformly in s ∈ (0, T]. For a Gaussian processesGfrom LSI(ρ), we consider a power variation V _n corresponding to a regular partition π _n of [0, T] and weighted by values of ρ(·). Under suitable hypotheses on G, we prove that a central limit theorem holds for V _n as the mesh of π _n approaches zero. The proof is based on a general central limit theorem for random variables that admit a Wiener chaos representation. The present result extends the central limit theorem for a power variation of a class of Gaussian processes with stationary increments and for bifractional and subfractional Gaussian processes.     

On the log-concavity of Hilbert series of Veronese subrings and Ehrhart series

For every positive integer n, consider the linear operator U _n on polynomials of degree at most d with integer coefficients defined as follows: if we write ${\frac{h(t)}{(1 - t)^{d + 1}}=\sum_{m \geq 0} g(m) \, t^{m}}For every positive integer n, consider the linear operator U _n on polynomials of degree at most d with integer coefficients defined as follows: if we write \frach(t)(1 - t)^{d + 1}=?_{m 3 0} g(m)  t^m{\frac{h(t)}{(1 - t)^{d + 1}}=\sum_{m \geq 0} g(m) \, t^{m}} , for some polynomial g(m) with rational coefficients, then \fracU_nh(t)(1- t)^d+1 = ?_{m 3 0}g(nm)  t^m{\frac{{\rm{U}}_{n}h(t)}{(1- t)^{d+1}} = \sum_{m \geq 0}g(nm) \, t^{m}} . We show that there exists a positive integer n _d , depending only on d, such that if h(t) is a polynomial of degree at most d with nonnegative integer coefficients and h(0) = 1, then for n ≥ n _d , U _n h(t) has simple, real, negative roots and positive, strictly log concave and strictly unimodal coefficients. Applications are given to Ehrhart δ-polynomials and unimodular triangulations of dilations of lattice polytopes, as well as Hilbert series of Veronese subrings of Cohen–Macauley graded rings.     

Maximizing a lower bound on the computational complexity

We solve a combinatorial problem that arises in determining by a method due to Engeler lower bounds for the computational complexity of algorithmic problems. Denote by G_d the class of permutation groups G of degree d that are iterated wreath products of symmetric groups, i.e. G = S_dh?11?1S_d0 with Π^h?1_i=0d_i = d for some natural number h and some sequence (d₀,…,d_h?1) of natural numbers greater than 1. The problem is to characterize those G = S_dh?11?1S_d0 in G_d on which k(G):=log|G|/max_0≤i≤h?1log(d_i!) assumes its maximum. Our solution consists of two necessary conditions for this, namely that d₀≤d₁≤?≤d_h and that d_h is the largest prime divisor of d. Consequently, if d is a prime power, say d = p^h with p prime, then a necessary and sufficient condition is that d_i = p, 0 ≤ i ≤ h ? 1.     

Tranchage et cône tangent des courants localement normaux

Let T be a locally normal current on an open set Ω of ℝ″ = ℝ x ℝ″⁻¹ and let π: ℝⁿ → ℝ denote the projection π(x₁, x″) = x₁. We define the current 〈T, π, 0〉 (called slice of T at 0 by π) as the limit, as ɛ → 0, of the family ɛ⁻¹TΛπ ^* (ψ(x¹ /ɛ)dx¹), where ψ is a C^∞ function on ℝ with compact support such that ∝ψ(x¹)dx¹ = 1, provided the limit exists and doesn't depend on the choice of ψ. We first prove that the limit lim_R→+∞(h_R)_#T exists, where h_R(x₁,x″) = (Rx¹,x″). We apply this result to the study of the existence of the tangent cone at 0 associated to a locally normal current, and especially associated to a subanalytic chain. We finally give a necessary and sufficient condition relative to T for the existence of the slice 〈T, π, 0〉.     

Defect indices of powers of a contraction

Let A be a contraction on a Hilbert space H. The defect index d_A of A is, by definition, the dimension of the closure of the range of I-A^∗A. We prove that (1) d_Aⁿ?nd_A for all n?0, (2) if, in addition, Aⁿ converges to 0 in the strong operator topology and d_A=1, then d_Aⁿ=n for all finite n,0?n?dimH, and (3) d_A=d_A^∗ implies d_Aⁿ=d_A^n∗ for all n?0. The norm-one index k_A of A is defined as sup{n?0:‖Aⁿ‖=1}. When dimH=m<∞, a lower bound for k_A was obtained before: k_A?(m/d_A)-1. We show that the equality holds if and only if either A is unitary or the eigenvalues of A are all in the open unit disc, d_A divides m and d_Aⁿ=nd_A for all n, 1?n?m/d_A. We also consider the defect index of f(A) for a finite Blaschke product f and show that d_f(A)=d_Aⁿ, where n is the number of zeros of f.     

On completeness of the systems {exp(ix(n+ihn))}

Рассматривается система функцийе(Λ)={ехр (ix(n+ih _n))}, где Imh _n=0, ±n=1, 2,... Если всеh _n=0, то эта система совпадает с тригонометрической системой, которая является неполной в С[?π, π]. Строится пример, показывающий, что в классе sup ¦h _n¦<∞ содержатся и полные вС[?π, π] системые(Λ). Доказывается, что условие sup¦h _n|=∞ уже влечет полноту системые(Λ) вС[?π, π], еслиh _n растут «не слишком быстро». Исследуется вопрос о степени переполненности системые(Λ).     

Metric Möbius geometry and a characterization of spheres

We obtain a Möbius characterization of the n-dimensional spheres S ⁿ endowed with the chordal metric d ₀. We show that every compact extended Ptolemy metric space with the property that every three points are contained in a circle is Möbius equivalent to (S ⁿ , d ₀) for some n ≥ 1.     

Homogeneous random fields and statistical mechanics

We illustrate the connection between homogeneous perturbations of homogeneous Gaussian random fields over Rⁿ or Zⁿ, with values in R^m, and classical as well as quantum statistical mechanics. In particular we construct homogeneous non-Gaussian random fields as weak limits of perturbed Gaussian random fields and study the infinite volume limit of correlation functions for a classical continuous gas of particles with inner degrees of freedom. We also exhibit the relation between quantum statistical mechanics of lattice systems (anharmonic crystals) at temperature β^?1 and homogeneous random fields over Zⁿ × S_β, where S_β is the circle of length β, which then provides a connection also with classical statistical mechanics. We obtain the infinite volume limit of real and imaginary times Green's functions and establish its properties. We also give similar results for the Gibbs state of the correspondent classical lattice systems and show that it is the limit as h → 0 of the quantum statistical Gibbs state.     

On boundary value problems for systems of nonlinear generalized ordinary differential equations

A general theorem (principle of a priori boundedness) on solvability of the boundary value problem dx = dA(t) · f(t, x), h(x) = 0 is established, where f: [a, b]×R ⁿ → R ⁿ is a vector-function belonging to the Carathéodory class corresponding to the matrix-function A: [a, b] → R ^n×n with bounded total variation components, and h: BV_s([a, b],R ⁿ ) → R ⁿ is a continuous operator. Basing on the mentioned principle of a priori boundedness, effective criteria are obtained for the solvability of the system under the condition x(t₁(x)) = B(x) · x(t ₂(x))+c ₀, where t _i: BV_s([a, b],R ⁿ ) → [a, b] (i = 1, 2) and B: BV_s([a, b], R ⁿ ) → R ⁿ are continuous operators, and c ₀ ∈ R ⁿ .     

On the equivalence of the centered Gaussian measure in L 2 with the correlation operator (−d 2/dx 2)−1 and the conditional Wiener measure

Let w and µ be respectively the conditional Wiener measure in C ₀([0, 1]) and the centered Gaussian measure in L ₂[0, 1] with the correlation operator (?d ²/dx ²)^?1. We prove the equivalence of these two measures in the following sense: for any Borel set A ? L ₂[0, 1] the set A ∩ C ₀([0, 1]) is a Borel subset of C ₀([0, 1]) and µ(A) = w(A∩C ₀([0, 1])).