Showing papers on "Function (mathematics) published in 2005"

Estimation of Non-Normalized Statistical Models by Score Matching

Abstract: One often wants to estimate statistical models where the probability density function is known only up to a multiplicative normalization constant. Typically, one then has to resort to Markov Chain Monte Carlo methods, or approximations of the normalization constant. Here, we propose that such models can be estimated by minimizing the expected squared distance between the gradient of the log-density given by the model and the gradient of the log-density of the observed data. While the estimation of the gradient of log-density function is, in principle, a very difficult non-parametric problem, we prove a surprising result that gives a simple formula for this objective function. The density function of the observed data does not appear in this formula, which simplifies to a sample average of a sum of some derivatives of the log-density given by the model. The validity of the method is demonstrated on multivariate Gaussian and independent component analysis models, and by estimating an overcomplete filter set for natural image data.

The Multiple-Sets Split Feasibility Problem and Its Applications for Inverse Problems

Abstract: The multiple-sets split feasibility problem requires finding a point closest to a family of closed convex sets in one space such that its image under a linear transformation will be closest to another family of closed convex sets in the image space. It can be a model for many inverse problems where constraints are imposed on the solutions in the domain of a linear operator as well as in the operator's range. It generalizes the convex feasibility problem as well as the two-sets split feasibility problem. We propose a projection algorithm that minimizes a proximity function that measures the distance of a point from all sets. The formulation, as well as the algorithm, generalize earlier work on the split feasibility problem. We offer also a generalization to proximity functions with Bregman distances. Application of the method to the inverse problem of intensity-modulated radiation therapy treatment planning is studied in a separate companion paper and is here only described briefly.

A Multimoment Bulk Microphysics Parameterization. Part II: A Proposed Three-Moment Closure and Scheme Description

Abstract: Many two-moment bulk schemes use a three-parameter gamma distribution of the form N(D) = N0Dαe−λD to describe the size spectrum of a given hydrometeor category. These schemes predict changes to the mass content and the total number concentration thereby allowing N0 and λ to vary as prognostic parameters while fixing the shape parameter, α. As was shown in Part I of this study, the shape parameter, which represents the relative dispersion of the hydrometeor size spectrum, plays an important role in the computation of sedimentation and instantaneous growth rates in bulk microphysics schemes. Significant improvement was shown by allowing α to vary as a diagnostic function of the predicted moments rather than using a fixed-value approach. Ideally, however, α should be an independent prognostic parameter. In this paper, a closure formulation is developed for calculating the source and sink terms of a third moment of the size distribution—the radar reflectivity. With predictive equations for the mass c...

The application of statistical physics to evolutionary biology.

Abstract: A number of fundamental mathematical models of the evolutionary process exhibit dynamics that can be difficult to understand analytically. Here we show that a precise mathematical analogy can be drawn between certain evolutionary and thermodynamic systems, allowing application of the powerful machinery of statistical physics to analysis of a family of evolutionary models. Analytical results that follow directly from this approach include the steady-state distribution of fixed genotypes and the load in finite populations. The analogy with statistical physics also reveals that, contrary to a basic tenet of the nearly neutral theory of molecular evolution, the frequencies of adaptive and deleterious substitutions at steady state are equal. Finally, just as the free energy function quantitatively characterizes the balance between energy and entropy, a free fitness function provides an analytical expression for the balance between natural selection and stochastic drift.

Regularity of Potential Functions of the Optimal Transportation Problem

Abstract: The potential function of the optimal transportation problem satisfies a partial differential equation of Monge-Ampere type. In this paper we introduce the notion of a generalized solution, and prove the existence and uniqueness of generalized solutions of the problem. We also prove the solution is smooth under certain structural conditions on the cost function.

Simulating Radiating and Magnetized Flows in Multi-Dimensions with ZEUS-MP

Abstract: This paper describes ZEUS-MP, a multi-physics, massively parallel, message- passing implementation of the ZEUS code. ZEUS-MP differs significantly from the ZEUS-2D code, the ZEUS-3D code, and an early "version 1" of ZEUS-MP distributed publicly in 1999. ZEUS-MP offers an MHD algorithm better suited for multidimensional flows than the ZEUS-2D module by virtue of modifications to the Method of Characteristics scheme first suggested by Hawley and Stone (1995), and is shown to compare quite favorably to the TVD scheme described by Ryu et. al (1998). ZEUS-MP is the first publicly-available ZEUS code to allow the advection of multiple chemical (or nuclear) species. Radiation hydrodynamic simulations are enabled via an implicit flux-limited radiation diffusion (FLD) module. The hydrodynamic, MHD, and FLD modules may be used in one, two, or three space dimensions. Self gravity may be included either through the assumption of a GM/r potential or a solution of Poisson's equation using one of three linear solver packages (conjugate-gradient, multigrid, and FFT) provided for that purpose. Point-mass potentials are also supported. Because ZEUS-MP is designed for simulations on parallel computing platforms, considerable attention is paid to the parallel performance characteristics of each module. Strong-scaling tests involving pure hydrodynamics (with and without self-gravity), MHD, and RHD are performed in which large problems (256^3 zones) are distributed among as many as 1024 processors of an IBM SP3. Parallel efficiency is a strong function of the amount of communication required between processors in a given algorithm, but all modules are shown to scale well on up to 1024 processors for the chosen fixed problem size.

More on the sum-product phenomenon in prime fields and its applications

Abstract: In this paper we establish new estimates on sum-product sets and certain exponential sums in finite fields of prime order. Our first result is an extension of the sum-product theorem from [8] when sets of different sizes are involed. It is shown that if and pe pδ (e)|A|. Next we exploit the Szemeredi–Trotter theorem in finite fields (also obtained in [8]) to derive several new facts on expanders and extractors. It is shown for instance that the function f(x,y) = x(x+y) from to satisfies |F(A,B)| > pβ for some β = β (α) > α whenever and |A| ~ |B|~ pα, 0 pρ where ρ pδ(e)|A|. This is the finite fields version of a problem considered in [11]. If A is an interval, there is a relation to estimates on incomplete Kloosterman sums. In the Appendix, we obtain an apparently new bound on bilinear Kloosterman sums over relatively short intervals (without the restrictions of Karatsuba's result [14]) which is of relevance to problems involving the divisor function (see [1]) and the distribution (mod p) of certain rational functions on the primes (cf. [12]).

Excessive Gap Technique in Nonsmooth Convex Minimization

Abstract: In this paper we introduce a new primal-dual technique for convergence analysis of gradient schemes for nonsmooth convex optimization. As an example of its application, we derive a primal-dual gradient method for a special class of structured nonsmooth optimization problems, which ensures a rate of convergence of order O(1/k), where k is the iteration count. Another example is a gradient scheme, which minimizes a nonsmooth strongly convex function with known structure with rate of convergence O(1/k2). In both cases the efficiency of the methods is higher than the corresponding black-box lower complexity bounds by an order of magnitude.

Information Bottleneck for Gaussian Variables

Abstract: The problem of extracting the relevant aspects of data was previously addressed through the information bottleneck (IB) method, through (soft) clustering one variable while preserving information about another - relevance - variable. The current work extends these ideas to obtain continuous representations that preserve relevant information, rather than discrete clusters, for the special case of multivariate Gaussian variables. While the general continuous IB problem is difficult to solve, we provide an analytic solution for the optimal representation and tradeoff between compression and relevance for the this important case. The obtained optimal representation is a noisy linear projection to eigenvectors of the normalized regression matrix Σx|yΣx-1, which is also the basis obtained in canonical correlation analysis. However, in Gaussian IB, the compression tradeoff parameter uniquely determines the dimension, as well as the scale of each eigenvector, through a cascade of structural phase transitions. This introduces a novel interpretation where solutions of different ranks lie on a continuum parametrized by the compression level. Our analysis also provides a complete analytic expression of the preserved information as a function of the compression (the "information-curve"), in terms of the eigenvalue spectrum of the data. As in the discrete case, the information curve is concave and smooth, though it is made of different analytic segments for each optimal dimension. Finally, we show how the algorithmic theory developed in the IB framework provides an iterative algorithm for obtaining the optimal Gaussian projections.

Smooth function approximation using neural networks

Abstract: An algebraic approach for representing multidimensional nonlinear functions by feedforward neural networks is presented. In this paper, the approach is implemented for the approximation of smooth batch data containing the function's input, output, and possibly, gradient information. The training set is associated to the network adjustable parameters by nonlinear weight equations. The cascade structure of these equations reveals that they can be treated as sets of linear systems. Hence, the training process and the network approximation properties can be investigated via linear algebra. Four algorithms are developed to achieve exact or approximate matching of input-output and/or gradient-based training sets. Their application to the design of forward and feedback neurocontrollers shows that algebraic training is characterized by faster execution speeds and better generalization properties than contemporary optimization techniques.

The Shape of Production Functions and the Direction of Technical Change

Abstract: This pai>er views the standard production function in macroeconomics as a reduced form and derives its properties from microfoundations. The shape of this production function is governed by the distribution of ideas. If that distribution is Pareto, then two results obtain: the global production function is Cobb-Douglas, and technical change in the long run is labor-augmenting. Kortum showed that Pareto distributions are necessary if search-based idea models are to exhibit steady-state growth. Here we show that this same assumption delivers the additional results about the shape of the production function and the direction of technical change.

Semilinear Duhem model for rate-independent and rate-dependent hysteresis

Abstract: The classical Duhem model provides a finite-dimensional differential model of hysteresis. In this paper, we consider rate-independent and rate-dependent semilinear Duhem models with provable properties. The vector field is given by the product of a function of the input rate and linear dynamics. If the input rate function is positively homogeneous, then the resulting input-output map of the model is rate independent, yielding persistent nontrivial input-output closed curve (that is, hysteresis) at arbitrarily low input frequency. If the input rate function is not positively homogeneous, the input-output map is rate dependent and can be approximated by a rate-independent model for low frequency inputs. Sufficient conditions for convergence to a limiting input-output map are developed for rate-independent and rate-dependent models. Finally, the reversal behavior and orientation of the rate-independent model are discussed.

The colored Jones function is q-holonomic

Abstract: A function of several variables is called holonomic if, roughly speaking, it is determined from finitely many of its values via finitely many linear recursion relations with polynomial coefficients. Zeilberger was the first to notice that the abstract notion of holonomicity can be applied to verify, in a systematic and computerized way, combinatorial identities among special functions. Using a general state sum definition of the colored Jones function of a link in 3-space, we prove from first principles that the colored Jones function is a multisum of a q-proper-hypergeometric function, and thus it is q-holonomic. We demonstrate our results by computer calculations.

Peel back user interface to show hidden functions

Abstract: A user is able to access additional functions not represented in a current image displayed by a graphical user interface. At least one function not presented on the current image is represented by a symbol on an underlying image that is at least partially covered by the current image. When the user performs a predetermined user input (e.g., selecting a corner of the current image), the underlying image and the at least one function represented thereby become accessible. When the user input is performed, a visual effect depicts the current image being at least partially removed from over the underlying image, thereby revealing and permitting access to the at least one additional function. The user input is made by the user performing an action with the user's hand or another object adjacent to a responsive display, or by using a pointing device to manipulate a displayed image.

The complex Monge-Ampère equation and pluripotential theory

Abstract: Positive currents and plurisubharmonic functions Siciak's extremal function and a related capacity The Dirichlet problem for the Monge-Ampere equation with continuous data The Dirichlet problem continued The Monge-Ampere equation for unbounded functions The complex Monge-Ampere equation on a compact Kahler manifold Bibliography.

Solar-like Oscillations in α Centauri B

Abstract: We have made velocity observations of the star α Centauri B from two sites, allowing us to identify 37 oscillation modes with l = 0-3. Fitting to these modes gives the large and small frequency separations as a function of frequency. The mode lifetime, as measured from the scatter of the oscillation frequencies about a smooth trend, is similar to that in the Sun. Limited observations of the star δ Pav show oscillations centered at 2.3 mHz, with peak amplitudes close to solar. We introduce a new method of measuring oscillation amplitudes from heavily smoothed power density spectra, from which we estimated amplitudes for α Cen α and B, β Hyi, δ Pav, and the Sun. We point out that the oscillation amplitudes may depend on which spectral lines are used for the velocity measurements.

Solar-like oscillations in alpha Centauri B

Abstract: We have made velocity observations of the star alpha Cen B from two sites, allowing us to identify 37 oscillation modes with l=0-3. Fitting to these modes gives the large and small frequency separations as a function of frequency. The mode lifetime, as measured from the scatter of the oscillation frequencies about a smooth trend, is similar to that in the Sun. Limited observations of the star delta Pav show oscillations centred at 2.3 mHz with peak amplitudes close to solar. We introduce a new method of measuring oscillation amplitudes from heavily-smoothed power density spectra, from which we estimated amplitudes for alpha Cen A and B, beta Hyi, delta Pav and the Sun. We point out that the oscillation amplitudes may depend on which spectral lines are used for the velocity measurements.

Fluctuation-dissipation theorem density-functional theory

Abstract: Using the fluctuation-dissipation theorem (FDT) in the context of density-functional theory (DFT), one can derive an exact expression for the ground-state correlation energy in terms of the frequency-dependent density response function. When combined with time-dependent density-functional theory, a new class of density functionals results that use approximations to the exchange-correlation kernel fxc as input. This FDT-DFT scheme holds promise to solve two of the most distressing problems of conventional Kohn–Sham DFT: (i) It leads to correlation energy functionals compatible with exact exchange, and (ii) it naturally includes dispersion. The price is a moderately expensive O(N6) scaling of computational cost and a slower basis set convergence. These general features of FDT-DFT have all been recognized previously. In this paper, we present the first benchmark results for a set of molecules using FDT-DFT beyond the random-phase approximation (RPA)—that is, the first such results with fxc≠0. We show that ke...

Efficient Collision-Resistant Hashing from Worst-Case Assumptions on Cyclic Lattices

Abstract: The generalized knapsack function is defined as f a (x)= Σ i a i x i , where a = (a 1 ,...,a m ) consists of m elements from some ring R, and x = (x i ,...,x m ) consists of m coefficients from a specified subset S C R. Micciancio (FOCS 2002) proposed a specific choice of the ring R and subset S for which inverting this function (for random a, x) is at least as hard as solving certain worst-case problems on cyclic lattices. We show that for a different choice of S ⊂ R, the generalized knapsack function is in fact collision-resistant, assuming it is infeasible to approximate the shortest vector in n-dimensional cyclic lattices up to factors 0(n). For slightly larger factors, we even get collision-resistance for any m > 2. This yields very efficient collision-resistant hash functions having key size and time complexity almost linear in the security parameter n. We also show that altering S is necessary, in the sense that Micciancio's original function is not collision-resistant (nor even universal one-way). Our results exploit an intimate connection between the linear algebra of n-dimensional cyclic lattices and the ring Z[α]/(α n -1), and crucially depend on the factorization of a n - 1 into irreducible cyclotomic polynomials. We also establish a new bound on the discrete Gaussian distribution over general lattices, employing techniques introduced by Micciancio and Regev (FOCS 2004) and also used by Micciancio in his study of compact knapsacks.

Reconstruction of solid models from oriented point sets

Abstract: In this paper we present a novel approach to the surface reconstruction problem that takes as its input an oriented point set and returns a solid, water-tight model. The idea of our approach is to use Stokes' Theorem to compute the characteristic function of the solid model (the function that is equal to one inside the model and zero outside of it). Specifically, we provide an efficient method for computing the Fourier coefficients of the characteristic function using only the surface samples and normals, we compute the inverse Fourier transform to get back the characteristic function, and we use iso-surfacing techniques to extract the boundary of the solid model.The advantage of our approach is that it provides an automatic, simple, and efficient method for computing the solid model represented by a point set without requiring the establishment of adjacency relations between samples or iteratively solving large systems of linear equations. Furthermore, our approach can be directly applied to models with holes and cracks, providing a method for hole-filling and zippering of disconnected polygonal models.

Airy Distribution Function: From the Area Under a Brownian Excursion to the Maximal Height of Fluctuating Interfaces

Abstract: The Airy distribution function describes the probability distribution of the area under a Brownian excursion over a unit interval. Surprisingly, this function has appeared in a number of seemingly unrelated problems, mostly in computer science and graph theory. In this paper, we show that this distribution function also appears in a rather well studied physical system, namely the fluctuating interfaces. We present an exact solution for the distribution P(h m ,L) of the maximal height h m (measured with respect to the average spatial height) in the steady state of a fluctuating interface in a one dimensional system of size L with both periodic and free boundary conditions. For the periodic case, we show that P(h m ,L)=L−1/2f(h m L−1/2) for all L>0 where the function f(x) is the Airy distribution function. This result is valid for both the Edwards–Wilkinson (EW) and the Kardar–Parisi–Zhang interfaces. For the free boundary case, the same scaling holds P(h m ,L)=L−1/2F(h m L−1/2), but the scaling function F(x) is different from that of the periodic case. We compute this scaling function explicitly for the EW interface and call it the F-Airy distribution function. Numerical simulations are in excellent agreement with our analytical results. Our results provide a rather rare exactly solvable case for the distribution of extremum of a set of strongly correlated random variables. Some of these results were announced in a recent Letter [S.N. Majumdar and A. Comtet, Phys. Rev. Lett. 92: 225501 (2004)].

Simulating biochemical networks at the particle level and in time and space: Green's function reaction dynamics.

Abstract: We present a technique, called Green's function reaction dynamics (GFRD), for particle-based simulations of reaction-diffusion systems. GFRD uses a maximum time step such that only single particles or pairs of particles have to be considered. For these particles, the Smoluchowski equations are solved analytically using Green's functions, which are used to set up an event-driven algorithm. We apply the technique to a model of gene expression. Under biologically relevant conditions, GFRD is up to 5 orders of magnitude faster than conventional particle-based schemes.

Nonlinear control of mechanical systems with an unactuated cyclic variable

Abstract: Numerous robotic tasks associated with underactuation have been studied in the literature. For a large number of these in the plane, the mechanical models have a cyclic variable, the cyclic variable is unactuated, and all shape variables are independently actuated. This paper formulates and solves two control problems for this class of models. If the generalized momentum conjugate to the cyclic variable is not conserved, conditions are found for the existence of a set of outputs that yields a system with a one-dimensional exponentially stable zero dynamics-i.e., an exponentially minimum-phase system-along with a dynamic extension that renders the system locally input-output decouplable. If the generalized momentum conjugate to the cyclic variable is conserved, a reduced system is constructed and conditions are found for the existence of a set of outputs that yields an empty zero dynamics, along with a dynamic extension that renders the system feedback linearizable. A common element in these two feedback problems is the construction of a scalar function of the configuration variables that has relative degree three with respect to one of the input components. The function arises by partially integrating the conjugate momentum. The results are illustrated on two balancing tasks and on a ballistic flip motion.

Model Selection for Regularized Least-Squares Algorithm in Learning Theory

Abstract: We investigate the problem of model selection for learning algorithms depending on a continuous parameter. We propose a model selection procedure based on a worst-case analysis and on a data-independent choice of the parameter. For the regularized least-squares algorithm we bound the generalization error of the solution by a quantity depending on a few known constants and we show that the corresponding model selection procedure reduces to solving a bias-variance problem. Under suitable smoothness conditions on the regression function, we estimate the optimal parameter as a function of the number of data and we prove that this choice ensures consistency of the algorithm.

A fast IE-FFT algorithm for solving PEC scattering problems

Abstract: This paper presents a novel fast integral equation method, termed IE-FFT, for solving large electromagnetic scattering problems Similar to other fast integral equation methods, the IE-FFT algorithm starts by partitioning the basis functions into multilevel clustering groups Subsequently, the entire impedance matrix is decomposed into two parts: one for the self and/or near field couplings, and one for well-separated group couplings The IE-FFT algorithm employs two discretizations one is for the unknown current on an unstructured triangular mesh, and the other is a uniform Cartesian grid for interpolating the Green's function By interpolating the Green's function on a regular Cartesian grid, the couplings between two well-separated groups can be computed using the fast Fourier transform (FFT) Consequently, the IE-FFT algorithm does not require the knowledge of addition theorem It simply utilizes the Toeplitz property of the coefficient matrix and is therefore applicable to both static and wave propagation problems