A theory for optimal regularization in the finite dimensional case

Consider the matrix problem $\text{Ax = y + ε =}\text{y}\text{?}$ in the case where A is known precisely, the problem is ill conditioned, and ε is a random noise vector. Compute regularized “ridge” estimates,$\text{x}\text{?}{}_{\mathrm{\lambda }}\text{= (A}{}^1\text{A + λI)}{}^{-1}\text{A}{}^1\text{y}\text{?}$,where ¹ denotes matrix transpose. Of great concern is the determination of the value of λ for which x?_λ “best” approximates $\text{x}{}_{\mathrm{<mtext>0</mtext>}}\text{= A}{}^{\mathrm{+}}\text{y}$. Let $\text{Q = 6}\text{x}\text{?}{}_{\mathrm{\lambda }}\text{? x}{}_{\mathrm{<mtext>0</mtext>}}\text{6}{}^2$,and define λ₀ to be the value of λ for which Q is a minimum. We look for λ₀ among solutions of dQ/dλ = 0. Though Q is not computable (since ε is unknown), we can use this approach to study the behavior of λ₀ as a function of y and ε. Theorems involving “noise to signal ratios” determine when λ₀ exists and define the cases λ₀ > 0 and λ₀ = ∞. Estimates for λ₀ and the minimum square error Q₀ = Q(λ₀) are derived.     

Indecomposable triple systems

A t-design (λ, t, d, n) is a system $\text{B}$ of sets of size d from an n-set S, such that each t subset of S is contained in exactly λ elements of $\text{B}$. A t-design is indecomposable (written IND(λ, t, d, n)) if there does not exist a subset $\text{B}$ ? $\text{B}$ such that $\text{B}$ is a (λ, t, d, n) for some λ, 1 ? λ < λ. A triple system is a (λ; 2, 3, n). Recursive and constructive methods (several due to Hanani) are employed to show that: (1) an IND(2; 2, 3, n) exists for n ≡ 0, 1 (mod 3), n ? 4 and n ≡ 7 (designs of Bhattacharya are used here), (2) an IND(3; 2, 3, n) exists for n odd, n ? 5, (3) if an IND(λ, 2, 3, n) exists, n odd, then there exists an infinite number of indecomposable triple systems with that λ.     

All directed BIBDs with k = 3 exist

A directed BIBD with parameters ($\text{υ, b, r, k, λ}{}^1$) is a BIBD with parameters ($\text{υ, b, r, k, 2λ}{}^1$) in which each ordered pair of varieties occurs together in exactly $\text{λ}{}^1$ blocks. It is shown that $\text{λ}{}^1\text{υ(υ ? 1) ≡ 0}$ (mod 3) is a necessary and sufficient condition for the existence of a directed ($\text{υ, b, r, k, λ}{}^1$) BIBD with k = 3.     

Perturbations with several independent parameters

In this paper we discuss the problem of determining a T-periodic solution $\text{x}{}^1\text{(·, λ)}$ of the differential equation x = A(t)x + f(t, x, λ) + b(t), where the perturbation parameter λ is a vector in a parameter-space R^k. The customary approach assumes that λ = λ(?), ??R. One then establishes the existence of an ?₀ > 0 such that the differential equation has a T-periodic solution $\text{x}{}^1\text{(·, λ(?))}$ for all ? satisfying 0 < ? < ?₀. More specifically it is usually assumed that λ(?) has the form λ(?) = ?λ₀ where λ₀ is a fixed vector in R^k. This means that attention is confined in the perturbation procedure to examining the dependence of $\text{x}{}^1\text{(·, λ)}$ on λ as λ varies along a line segment terminating at the origin in the parameter-space R^k. The results established here generalize this previous work by allowing one to study the dependence of $\text{x}{}^1\text{(·, λ)}$ on λ as λ varies through a “conical-horn” whose vertex rests at the origin in R^k. In the process an implicit-function formula is developed which is of some interest in its own right.     

Z3-invariants of real and imaginary quadratic fields

Let $\text{Q}$((?m)^{$\text{1}\text{2}$}) and $\text{Q}$((3m)^{$\text{1}\text{2}$}) be a pair of quadratic fields, m > 0, and let λ^?, μ^?; λ⁺, μ⁺ be the respective Iwasawa invariants of the basic $\text{Z}$₃-extensions of these fields. A generalization of a result of Scholz shows that λ^? ≥ λ⁺ and if μ^? = 0, then μ⁺ = 0.     

Cantor spectrum for the almost Mathieu equation

For a dense G_δ of pairs (λ, α) in R², we prove that the operator (Hu)(n) = u(n + 1) + u(n ?1) + λ cos(2παn + θ) u(n) has a nowhere dense spectrum. Along the way we prove several interesting results about the case $\text{α =}\text{p}\text{q}$ of which we mention: (a) If qθ is not an integral multiple of π, then all gaps are open, and (b) If q is even and θ is chosen suitably, then the middle gap is closed for all λ.     

The chromatic polynomial of an unlabeled graph

We investigate the chromatic polynomial χ(G, λ) of an unlabeled graph G. It is shown that $\text{χ(G, λ) = (}\text{1}\text{|A(g)|}\text{) Σ}{}_{\mathrm{\pi \; \in \; A(g)}}\text{χ(g, π, λ)}$, where g is any labeled version of G, A(g) is the automorphism group of g and χ(g, π, λ) is the chromatic polynomial for colorings of g fixed by π. The above expression shows that χ(G, λ) is a rational polynomial of degree n = |V(G)| with leading coefficient $\text{1}\text{|A(g)|}$. Though χ(G, λ) does not satisfy chromatic reduction, each polynomial χ(g, π, λ) does, thus yielding a simple method for computing χ(G, λ). We also show that the number N(G) of acyclic orientations of G is related to the argument λ = ?1 by the formula $\text{N(G) = (}\text{1}\text{|A(g)|}\text{) Σ}{}_{\mathrm{\pi \; \in \; A(g)}}\text{(?1)}{}^{\mathrm{s(\pi )}}\text{χ(g, π, ?1)}$, where s(π) is the number of cycles of π. This information is used to derive Robinson's (“Combinatorial Mathematics V” (Proc. 5th Austral. Conf. 1976), Lecture Notes in Math. Vol. 622, pp. 28–43, Springer-Verlag, New York/Berlin, 1977) cycle index sum equations for counting unlabeled acyclic digraphs.     

On the maximum number of permutations with given maximal or minimal distance

Let us denote by R(k, ? λ)[R(k, ? λ)] the maximal number $\text{M}$ such that there exist $\text{M}$ different permutations of the set {1,…, k} such that any two of them have at least λ (at most λ, respectively) common positions. We prove the inequalities R(k, ? λ) ? kR(k ? 1, ? λ ? 1), R(k, ? λ) ? R(k, ? λ ? 1) ? k!, R(k, ? λ) ? kR(k ? 1, ? λ ? 1). We show: R(k, ? k ? 2) = 2, R(k, ? 1) = (k ? 1)!, R(p^m, ? 2) = (p^m ? 2)!, R(p^m + 1, ? 3) = (p^m ? 2)!, $\text{R(k, ? k ? 3) =}\text{k!}\text{2}$, R(k, ? 0) = k, R(p^m, ? 1) = p^m(p^m ? 1), R(p^m + 1, ? 2) = (p^m + 1)p^m(p^m ? 1). The exact value of R(k, ? λ) is determined whenever k ? k₀(k ? λ); we conjecture that R(k, ? λ) = (k ? λ)! for k ? k₀(λ). Bounds for the general case are given and are used to determine that the minimum of |R(k, ? λ) ? R(k, ? λ)| is attained for $\text{λ = (}\text{k}\text{2}\text{) + O(}\text{k}\text{log}\text{k}\text{)}$.     

On the absorption of eigenvalues by continuous spectrum in regular perturbation problems

We consider the family of operators A + λB with A and B self-adjoint and B relatively form bounded. We consider situations where as λ ↓ λ₁, some eigenvalue μ(λ) approaches the continuous spectrum of A + λB. Typical of our results is the following. If B is relatively form compact, and μ(λ) → μ(λ₁), then either $\text{(μ(λ) ? μ(λ}{}_1\text{))}\text{λ}\text{? λ}{}_1\text{→ 0}\text{or}\text{μ(λ}{}_1\text{)}$ is an eigenvalue of A + λ₁B.     

Multiple solutions of coupled Sturm-Liouville systems

This paper deals with the coupled Sturm-Liouville system ? (pu′)′ + Pu + rv = λ₁u + λ₁N₁₁(u, v) + λ₂N₂₁(u, v), ? (qv′)′ + Qv + ru = λ₂v + λ₁N₁₂(u, v) + λ₂N₂₂(u, v), α₁₁u(0) + α₁₂u′(0) = 0 = α₂₁v(0) + α₂₂v′(0), β₁₁u(1) + β₁₂u′(1) = 0 = β₂₁v(1) + β₂₂v′(1). The functions p, P, q, Q, r are smooth; λ₁ and λ₂ are eigenparameters; N_ij(u, v) is analytic and of higher order. The linearized problem, all $\text{N}{}_{\mathrm{ij}}\text{\&z.tbnd; 0}$, is shown to have eigenvalues (λ₁, λ₂) which are continuously distributed along a sequence of monotonically decreasing curves in the λ₁λ₂-plane. A generalized Lyapunov-Schmidt method establishes that if (λ₁, λ₂) is near a simple eigenvalue of the linearized problem, then the number of small solutions of the nonlinear problem corresponds to the number of real roots of a certain polynomial.     

Finite-amplitude disturbances in self-gravitating media. II

This paper deals with finite-amplitude axisymmetric disturbances in a self-gravitating fluid column of finite radius R. It is shown that the cutoff wavelength λ_nl above which gravitational breakup occurs now depends on the relative amplitude $\text{?}\text{R}$ of the initial perturbation. Actually, for small-but finite-amplitude disturbances, $\text{λ}{}_{\mathrm{nl}}\text{=}\text{λ}{}_{\mathrm{l}}\text{(1 ? 0.34368}\text{?}{}^2\text{R}{}^2\text{)}$, where λ_l ( = 5.8898R) designates the cutoff wavelength predicted in the linear approximation.     

Sturm—Liouville problems with several parameters

We consider the regular linear Sturm-Liouville problem (second-order linear ordinary differential equation with boundary conditions at two points x = 0 and x = 1, those conditions being separated and homogeneous) with several real parameters λ₁,…,λ_N. Solutions to this problem correspond to eigenvaluesλ = (λ₁,…,λ_N) forming sets $\text{R}$^N determined by the number of zeroes in (0, 1) of solutions. We describe properties of these sets including: boundedness, and when unbounded, asymptotic directions. Using these properties some results are given for the system of N Sturm-Liouville problems which share only the parameters λ. Sharp results are given for the system of two problems sharing two parameters. The eigensurfaces for a single problem are closely related to the cone $\text{K={λ R}{}^{\mathrm{N}}\text{:λ}{}_1\text{a}{}_1\text{(x)+…+λ}{}_{\mathrm{N}}\text{a}{}_{\mathrm{N}}\text{(x)?0 for all x in [0,1]}}$, particularly in questions of boundedness. The cone K and related objects are discussed, and a result is given which relates cones with two oscillation conditions known as “Right-Definiteness” and “Left-Definiteness.”     

A maximization problem and its application to canonical correlation

Let Σ be an n × n positive definite matrix with eigenvalues λ₁ ≥ λ₂ ≥ … ≥ λ_n > 0 and let M = {x, y | x?Rⁿ, y?Rⁿ, x ≠ 0, y ≠ 0, x′y = 0}. Then for x, y in M, we have that $\text{x′Σy}\text{(x′Σxy′Σy)}{}^{\mathrm{<mtext>1</mtext><mtext>2</mtext>}}\text{≤}\text{(λ}{}_1\text{? λ}{}_{\mathrm{n}}\text{)}\text{(λ}{}_1\text{+ λ}{}_{\mathrm{n}}\text{)}$ and the inequality is sharp. If

$\text{∑=}\text{∑}{}_{11}\text{∑}{}_{12}\text{∑}{}_{21}\text{∑}{}_{22}$

is a partitioning of Σ, let θ₁ be the largest canonical correlation coefficient. The above result yields $\text{θ}{}_1\text{≤}\text{(λ}{}_1\text{? λ}{}_{\mathrm{n}}\text{)}\text{(λ}{}_1\text{+ λ}{}_{\mathrm{n}}\text{)}$.     

An example of bifurcation to homoclinic orbits

Consider the equation $\text{x}\text{?}\text{? x + x}{}^2\text{= ?λ}{}_1\text{x + λ}{}_2\text{?(t)}\text{where}\text{?(t + 1) = ?(t)}\text{and}\text{λ = (λ}{}_1\text{, λ}{}_2\text{)}$ is small. For λ = 0, there is a homoclinic orbit Γ through zero. For λ ≠ 0 and small, there can be “strange” attractors near Γ. The purpose of this paper is to determine the curves in λ-space of bifurcation to “strange” attractors and to relate this to hyperbolic subharmonic bifurcations.     

On relative primeness of operator polynomials

Given operator polynomials A(λ) and B(λ), one of which is monic, conditions are given for the existence of operator polynomials C(λ) and D(λ) such that A(λ)C(λ) + B(λ)D(λ) = I for every $\text{λ ∈}\text{C}$. A special case will give a characterization of controlla- bility of an infinite-dimensional linear control system.     

A fundamental region for Hecke's modular group

Hecke proved analytically that when λ ≥ 2 or when $\text{λ = 2}\text{cos}\text{(}\text{π}\text{q}\text{)}$, q ∈ Z, q ≥ 3, then $\text{B(λ) = {τ:}\text{Im}\text{τ > 0, |}\text{Re}\text{τ| <}\text{λ}\text{2}\text{, |τ| > 1}}$ is a fundamental region for the group G(λ) = 〈S_λ, T〉, where S_λ: τ → τ + λ and $\text{T: τ →}\text{?1}\text{τ}$. He also showed that B(λ) fails to be a fundamental region for all other λ > 0 by proving that G(λ) is not discontinuous. We give an elementary proof of these facts and prove a related result concerning the distribution of G(λ)-equivalent points.     

Nonlinear eigenvalue problems with positively convex operators

We consider the equation u = λAu (λ > 0), where A is a forced isotone positively convex operator in a partially ordered normed space with a complete positive cone K. Let Λ be the set of positive λ for which the equation has a solution u?K, and let Λ₀ be the set of positive λ for which a positive solution—necessarily the minimum one—can be obtained by an iteration u_n = λAu_n?1, u₀ = 0. We show that if K is normal, and if Λ is nonempty, then Λ₀ is nonempty, and each set Λ₀, Λ is an interval with $\text{inf}\text{(Λ}{}_0\text{) =}\text{inf}\text{(Λ) = 0}\text{and sup}\text{(Λ}{}_0\text{) =}\text{sup}\text{(Λ) (= λ1,}\text{say}\text{)}$; but we may have λ1 ? Λ₀ and λ1 ? Λ. Furthermore, if A is bounded on the intersection of K with a neighborhood of 0, then Λ₀ is nonempty. Let u⁰(λ) = lim_n→∞(λA)ⁿ(0) be the minimum positive fixed point corresponding to λ ? Λ₀. Then u⁰(λ) is a continuous isotone convex function of λ on Λ₀.     

Resolvable perfect cyclic designs

Let k, λ, and υ be positive integers. A perfect cyclic design in the class PD(υ, k, λ) consists of a pair (Q, B) where Q is a set with |Q| = υ and B is a collection of cyclically ordered k-subsets of Q such that every ordered pair of elements of Q are t apart in exactly λ of the blocks for t = 1, 2, 3,…, k?1. To clarify matters the block [a₁, a₂, …, a_k] has cyclic order a₁ < a₂ < a₃ … < a_k < a₁ and a_i and a_i+1 are said to be t apart in the block where i + t is taken mod k. In this paper we are interested only in the cases where λ = 1 and υ ≡ 1 mod k. Such a design has $\text{υ(υ ? 1)}\text{k}$ blocks. If the blocks can be partitioned into υ sets containing $\text{(υ ? 1)}\text{k}$ pairwise disjoint blocks the design is said to be resolvable, and any such partitioning of the blocks is said to be a resolution. Any set of $\text{υ ? 1)}\text{k}$ pairwise disjoint blocks together with a singleton consisting of the only element not in one of the blocks is called a parallel class. Any resolution of a design yields υ parallel classes. We denote by RPD(υ, k, 1) the class of all resolvable perfect cyclic designs with parameters υ, k, and 1. Associated with any resolvable perfect cyclic design is an orthogonal array with k + 1 columns and υ rows with an interesting conjugacy property. Also a design in the class RPD(υ, k, 1) is constructed for all sufficiently large υ with υ ≡ 1 mod k.     

Quaternionic compositum genus

A quaternionic field over the rationals contains three quadratic subfields with a compositum genus relation of the type described in the author's paper in Volume 9 of this journal, involving the representation of a prime as norm in these subfields. These representations had previously been only partially exlored by the transfer of class structure from the rational to the quadratic fields. Here a full exposition is given by constructing the Artin characters when the subfields are $\text{Q}$(2^1/2), $\text{Q}$(q^1/2), and $\text{Q}$(2q)^1/2 (q prime). A special role belongs to q = A² + 32b².     

On the convergence of solutions of functional differential equations with infinite delay

Let $\text{C(β), S1(β),}\text{and}\text{K(β, λ)}$ be the classes of univalent functions defined in $\text{E = {z: ¦z¦< 1}}$, which are convex of order β, starlike of order β and close-to-convex of order β type λ. Let $\text{f(z) = (}\text{1}\text{α}\text{)z}{}^{\mathrm{<mtext>1?</mtext><mtext>1</mtext><mtext>\alpha </mtext>}}\text{∝}{}^{\mathrm{z}}{}_0\text{z}{}^{\mathrm{<mtext>1</mtext><mtext>x?2</mtext>}}\text{F(z)dz, 0 ? α < 1}$. We discuss the properties of the function f when this function F belongs to the class K(β, λ) and its various subclasses.