Exact Arithmetic Research Articles

Generalizing Random Butterfly Transforms to Arbitrary Matrix Sizes

Parker and Lê introduced random butterfly transforms (RBTs) as a preprocessing technique to replace pivoting in dense LU factorization. Unfortunately, their FFT-like recursive structure restricts the dimensions of the matrix. Furthermore, on multi-node systems, efficient management of the communication overheads restricts the matrix’s distribution even more. To remove these limitations, we have generalized the RBT to arbitrary matrix sizes by truncating the dimensions of each layer in the transform. We expanded Parker’s theoretical analysis to generalized RBT, specifically that in exact arithmetic, Gaussian elimination with no pivoting will succeed with probability 1 after transforming a matrix with full-depth RBTs. Furthermore, we experimentally show that these generalized transforms improve performance over Parker’s formulation by up to 62 % while retaining the ability to replace pivoting. This generalized RBT is available in the SLATE numerical software library.

Affective and cognitive factors associated with Chinese and Italian children’s arithmetic performance

BackgroundThis study aimed to investigate the cognitive and affective factors associated with cross-cultural differences in arithmetic tasks.MethodsA total of 404 third- and fourth- graders were recruited from China and Italy to complete exact arithmetic, arithmetic estimation and cognitive tasks (i.e., short-term memory, executive functions, and fluid reasoning). Their mathematical anxiety was also measured.ResultsThe results showed that Chinese children performed better than Italian children in both arithmetic tasks and in shifting task. Italian children performed better in visuospatial updating task and reported higher levels of mathematical anxiety than their Chinese peers. Multi-group path analyses showed that the patterns of relations among cognitive factors (i.e., short-term memory, inhibition and shifting), mathematical anxiety, and arithmetic performance were similar across groups. The only exception was that visuospatial updating uniquely predicted arithmetic estimation for Chinese but not for Italian children.ConclusionsChinese children outperformed their Italian peers in the exact arithmetic task, likely due to the greater emphasis on arithmetic fluency in Chinese mathematics education, both in schools and at home. They also had a slight advantage than Italian peers in the arithmetic estimation task. The unique link between updating and arithmetic estimation found in Chinese children but not Italian children suggests that, although arithmetic estimation is not emphasized in the curricula of either country, instruction and practice in exact arithmetic may enhance Chinese children’s efficiency in solving arithmetic estimation problems.

Global Convergence of Hessenberg Shifted QR I: Exact Arithmetic

Global Convergence of Hessenberg Shifted QR I: Exact Arithmetic

Relaxations and Exact Solutions to Quantum Max Cut via the Algebraic Structure of Swap Operators

The Quantum Max Cut (QMC) problem has emerged as a test-problem for designing approximation algorithms for local Hamiltonian problems. In this paper we attack this problem using the algebraic structure of QMC, in particular the relationship between the quantum max cut Hamiltonian and the representation theory of the symmetric group.The first major contribution of this paper is an extension of non-commutative Sum of Squares (ncSoS) optimization techniques to give a new hierarchy of relaxations to Quantum Max Cut. The hierarchy we present is based on optimizations over polynomials in the qubit swap operators. This is in contrast to the "standard" quantum Lasserre Hierarchy, which is based on polynomials expressed in terms of the Pauli matrices. To prove correctness of this hierarchy, we exploit a finite presentation of the algebra generated by the qubit swap operators. This presentation allows for the use of computer algebraic techniques to manipulate and simplify polynomials written in terms of the swap operators, and may be of independent interest. Surprisingly, we find that level-2 of this new hierarchy is numerically exact (up to tolerance 10−7) on all QMC instances with uniform edge weights on graphs with at most 8 vertices.The second major contribution of this paper is a polynomial-time algorithm that computes (in exact arithmetic) the maximum eigenvalue of the QMC Hamiltonian for certain graphs, including graphs that can be "decomposed" as a signed combination of cliques. A special case of the latter are complete bipartite graphs with uniform edge-weights, for which exact solutions are known from the work of Lieb and Mattis \cite{lieb1962ordering}. Our methods, which use representation theory of the symmetric group, can be seen as a generalization of the Lieb-Mattis result.

Chaos in Stochastic 2d Galerkin-Navier–Stokes

We prove that all Galerkin truncations of the 2d stochastic Navier–Stokes equations in vorticity form on any rectangular torus subjected to hypoelliptic, additive stochastic forcing are chaotic at sufficiently small viscosity, provided the frequency truncation satisfies N≥392\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$N\\ge 392$$\\end{document}. By “chaotic” we mean having a strictly positive Lyapunov exponent, i.e. almost-sure asymptotic exponential growth of the derivative with respect to generic initial conditions. A sufficient condition for such results was derived in previous joint work with Alex Blumenthal which reduces the question to the non-degeneracy of a matrix Lie algebra implying Hörmander’s condition for the Markov process lifted to the sphere bundle (projective hypoellipticity). The purpose of this work is to reformulate this condition to be more amenable for Galerkin truncations of PDEs and then to verify this condition using (a) a reduction to genericity properties of a diagonal sub-algebra inspired by the root space decomposition of semi-simple Lie algebras and (b) computational algebraic geometry executed by Maple in exact rational arithmetic. Note that even though we use a computer assisted proof, the result is valid for all aspect ratios and all sufficiently high dimensional truncations; in fact, certain steps simplify in the formal infinite dimensional limit.

Robust Coarse Cage Construction With Small Approximation Errors.

We propose a robust and automatic method to construct manifold cages for 3D triangular meshes. The cage contains hundreds of triangles to tightly enclose the input mesh without self-intersections. To generate such cages, our algorithm consists of two phases: (1) construct manifold cages satisfying the tightness, enclosing, and intersection-free requirements and (2) reduce mesh complexities and approximation errors without violating the enclosing and intersection-free requirements. To theoretically make the first stage have those properties, we combine the conformal tetrahedral meshing and tetrahedral mesh subdivision. The second step is a constrained remeshing process using explicit checks to ensure that the enclosing and intersection-free constraints are always satisfied. Both phases use a hybrid coordinate representation, i.e., rational numbers and floating point numbers, combined with exact arithmetic and floating point filtering techniques to guarantee the robustness of geometric predicates with a favorable speed. We extensively test our method on a data set of over 8500 models, demonstrating robustness and performance. Compared to other state-of-the-art methods, our method possesses much stronger robustness.

Pure quadratization and solution of ordinary differential equations by probabilistic evolution theory with concurrent computation of coefficients using exact arithmetic

Pure quadratization and solution of ordinary differential equations by probabilistic evolution theory with concurrent computation of coefficients using exact arithmetic

A Domain-theoretic Approach to Statistical Programming Languages

We give a domain-theoretic semantics to a statistical programming language, using the plain old category of dcpos, in contrast to some more sophisticated recent proposals. Remarkably, our monad of minimal valuations is commutative, which allows for program transformations that permute the order of independent random draws, as one would expect. A similar property is not known for Jones and Plotkin’s monad of continuous valuations. Instead of working with true real numbers, we work with exact real arithmetic, providing a bridge towards possible implementations (implementations by themselves are not addressed here). Rather remarkably, we show that restricting ourselves to minimal valuations does not restrict us much: All measures on the real line can be modeled by minimal valuations on the domain I ℝ ⊥ of exact real arithmetic. We give three operational semantics for our language, and we show that they are all adequate with respect to the denotational semantics. We also explore quite a few examples to demonstrate that our semantics computes exactly as one would expect and to debunk the myth that a semantics based on continuous maps would not be expressive enough to encode measures with non-compact support using only measures with compact support, or to encode measures via non-continuous density functions, for instance. Our examples also include some useful, non-trivial cases of distributions on higher-order objects.

Algebraic number fields and the LLL algorithm

In this paper we analyze the computational costs of various operations and algorithms in algebraic number fields using exact arithmetic. Let K be an algebraic number field. In the first half of the paper, we calculate the running time and the size of the output of many operations in K in terms of the size of the input and the parameters of K. We include some earlier results about these, but we go further than them, e.g. we also analyze some R-specific operations in K like less-than comparison.In the second half of the paper, we analyze two algorithms: the Bareiss algorithm, which is an integer-preserving version of the Gaussian elimination, and the LLL algorithm, which is for lattice basis reduction. In both cases, we extend the algorithm from Zn to Kn, and give a polynomial upper bound on the running time when the computations in K are performed exactly (as opposed to floating-point approximations).

The development of computational estimation strategies: a longitudinal study with 6- to 7-year-olds

Research showed that the capacity of making simple estimations begins to develop already at the age of five, but little is known about the early development of this estimation capacity and the strategies that underly it. The current study longitudinally followed the estimation capacity and strategies of 332 children from first to second grade of primary school by gathering strategy information through immediate trial-by-trial verbal reports on 14 addition items divided over six number size levels. Both in first and second grade children already used strategies that exhibited a basic understanding of computational estimation. For instance, their verbalizations suggested that they knew that a good estimate had to be close to the exact answer without necessarily coinciding with it. In addition to these rudimentary computational estimation strategies, we observed an increasing urge in those children to solve computational estimation problems by means of exact arithmetic. Implications for the earlier introduction of computational estimation in primary mathematics education are discussed.

Rational QZ steps with perfect shifts

AbstractIn this paper we analyze the stability of the problem of performing a rational QZ step with a shift that is an eigenvalue of a given regular pencil $$H-\lambda K$$ H - λ K in unreduced Hessenberg–Hessenberg form. In exact arithmetic, the backward rational QZ step moves the eigenvalue to the top of the pencil, while the rest of the pencil is maintained in Hessenberg–Hessenberg form, which then yields a deflation of the given shift. But in finite-precision the rational QZ step gets “blurred” and precludes the deflation of the given shift at the top of the pencil. In this paper we show that when we first compute the corresponding eigenvector to sufficient accuracy, then the rational QZ step can be constructed using this eigenvector, so that the exact deflation is also obtained in finite-precision.

Newton's method revisited: How accurate do we have to be?

SummaryWe analyze the convergence of quasi‐Newton methods in exact and finite precision arithmetic using three different techniques. We derive an upper bound for the stagnation level and we show that any sufficiently exact quasi‐Newton method will converge quadratically until stagnation. In the absence of sufficient accuracy, we are likely to retain rapid linear convergence. We confirm our analysis by computing square roots and solving bond constraint equations in the context of molecular dynamics. In particular, we apply both a symmetric variant and Forsgren's variant of the simplified Newton method. This work has implications for the implementation of quasi‐Newton methods regardless of the scale of the calculation or the machine.

Fast floating-point filters for robust predicates

Geometric predicates are at the core of many algorithms, such as the construction of Delaunay triangulations, mesh processing and spatial relation tests. These algorithms have applications in scientific computing, geographic information systems and computer-aided design. With floating-point arithmetic, these geometric predicates can incur round-off errors that may lead to incorrect results and inconsistencies, causing computations to fail. This issue has been addressed using a combination of exact arithmetic for robustness and floating-point filters to mitigate the computational cost of exact computations. The implementation of exact computations and floating-point filters can be a difficult task, and code generation tools have been proposed to address this. We present a new C++ meta-programming framework for the generation of fast, robust predicates for arbitrary geometric predicates based on polynomial expressions. We combine and extend different approaches to filtering, branch reduction, and overflow avoidance that have previously been proposed. We show examples of how this approach produces correct results for data sets that could lead to incorrect predicate results with naive implementations. Our benchmark results demonstrate that our implementation surpasses state-of-the-art implementations.

Characterizing exact arithmetic abilities before formal schooling

Children appear to have some arithmetic abilities before formal instruction in school, but the extent of these abilities as well as the mechanisms underlying them are poorly understood. Over two studies, an initial exploratory study of preschool children in the U.S. (N = 207; Age = 2.89–4.30 years) and a pre-registered replication of preschool children in Italy (N = 130; Age = 3–6.33 years), we documented some basic behavioral signatures of exact arithmetic using a non-symbolic subtraction task. Furthermore, we investigated the underlying mechanisms by analyzing the relationship between individual differences in exact subtraction and assessments of other numerical and non-numerical abilities. Across both studies, children performed above chance on the exact non-symbolic arithmetic task, generally showing better performance on problems involving smaller quantities compared to those involving larger quantities. Furthermore, individual differences in non-verbal approximate numerical abilities and exact cardinal number knowledge were related to different aspects of subtraction performance. Specifically, non-verbal approximate numerical abilities were related to subtraction performance in older but not younger children. Across both studies we found evidence that cardinal number knowledge was related to performance on subtraction problems where the answer was zero (i.e., subtractive negation problems). Moreover, subtractive negation problems were only solved above chance by children who had a basic understanding of cardinality. Together these finding suggest that core non-verbal numerical abilities, as well as emerging knowledge of symbolic numbers provide a basis for some, albeit limited, exact arithmetic abilities before formal schooling.

A New Bound for Hadwiger’s Covering Problem in \(\boldsymbol{\mathbb{E}}^3\)

We show that every three-dimensional convex body can be covered by 14 smaller homothetic copies. The previous result was 16 established by Papadoperakis in 1999, while a conjecture by Hadwiger is 8. We develop a discretization technique showing that it suffices to verify the feasibility of a number of linear programs with rational coefficients, which is done with computer assistance using exact arithmetic.

Lookahead analysis in exact real arithmetic with logical methods

Program extraction from proofs can be used to obtain verified algorithms in exact real arithmetic for e.g., the signed-digit code representation. In Minlog this has been done in the past with the use of certain coinductive predicates. In a next step we want to analyze the lookahead of these extracted programs. Doing it by hand is quite cumbersome, so instead we change our definition. Instead of a coinductive predicate we use an inductive predicate for the representation of reals that already incorporates the lookahead. In this way the lookahead becomes part of the specification which is carried through all the proofs. In the end we extract programs for computations on a signed-digit representation and we can just read off the lookahead from the specification.

On the backward and forward error of approximations of analytic functions and applications to the computation of matrix functions

In this paper we give a new formula to write the forward error of Taylor approximations of analytical functions in terms of the backward error of those approximations, considering exact arithmetic in both errors. Using this formula, a method to compute a backward error given by the power series centered in the same expansion point as the Taylor approximation is provided. The application of this method for Padé approximations is also shown. Based on the previous method, a MATLAB implementation for computing the first power series terms of the backward error for Taylor and Padé approximations of a given analytical function is provided, and examples of its use are given. Applications to the computation of matrix functions are given that overcome limitations of other backward error analyses use inverse compositional functions in the literature.

Algorithm 1021: SPEX Left LU, Exactly Solving Sparse Linear Systems via a Sparse Left-looking Integer-preserving LU Factorization

SPEX Left LU is a software package for exactly solving unsymmetric sparse linear systems. As a component of the sparse exact (SPEX) software package, SPEX Left LU can be applied to any input matrix, A , whose entries are integral, rational, or decimal, and provides a solution to the system \( Ax = b \) , which is either exact or accurate to user-specified precision. SPEX Left LU preorders the matrix A with a user-specified fill-reducing ordering and computes a left-looking LU factorization with the special property that each operation used to compute the L and U matrices is integral. Notable additional applications of this package include benchmarking the stability and accuracy of state-of-the-art linear solvers and determining whether singular-to-double-precision matrices are indeed singular. Computationally, this article evaluates the impact of several novel pivoting schemes in exact arithmetic, benchmarks the exact iterative solvers within Linbox, and benchmarks the accuracy of MATLAB sparse backslash. Most importantly, it is shown that SPEX Left LU outperforms the exact iterative solvers in run time on easy instances and in stability as the iterative solver fails on a sizeable subset of the tested (both easy and hard) instances. The SPEX Left LU package is written in ANSI C, comes with a MATLAB interface, and is distributed via GitHub, as a component of the SPEX software package, and as a component of SuiteSparse.

Error Bounds for Lanczos-Based Matrix Function Approximation

We analyze the Lanczos method for matrix function approximation (Lanczos-FA), an iterative algorithm for computing $f(\mathbf{A}) \mathbf{b}$ when $\mathbf{A}$ is a Hermitian matrix and $\mathbf{b}$ is a given vector. Assuming that $f : \mathbb{C} \rightarrow \mathbb{C}$ is piecewise analytic, we give a framework, based on the Cauchy integral formula, which can be used to derive a priori and a posteriori error bounds for Lanczos-FA in terms of the error of Lanczos used to solve linear systems. Unlike many error bounds for Lanczos-FA, these bounds account for fine-grained properties of the spectrum of $\mathbf{A}$, such as clustered or isolated eigenvalues. Our results are derived assuming exact arithmetic, but we show that they are easily extended to finite precision computations using existing theory about the Lanczos algorithm in finite precision. We also provide generalized bounds for the Lanczos method used to approximate quadratic forms $\mathbf{b}^\textsf{H} f(\mathbf{A}) \mathbf{b}$, and demonstrate the effectiveness of our bounds with numerical experiments.

Improving mathematics performance in 7-year-old children: Training the mapping from estimated quantities to Arabic digits

Exact arithmetic abilities require symbolic numerals, which constitute a precise representation of quantities, such as the Arabic digits. Numerical thinking, however, also engages an intuitive non-linguistic number sense, the Approximate Number System (ANS). The ANS allows us to discriminate quantities, approximate arithmetic transformations, and estimate quantities, all without counting individual items. Although the ANS does not require language, estimations made by means of the ANS can be expressed with number words or Arabic digits. A connection between the ANS and school math performance has been established. A child’s accuracy in mapping from approximate quantities to Arabic digits is associated with children’s symbolic math abilities and can also predict their success at learning new arithmetic skills. Here, we explore whether directly training the mapping between estimated quantities and Arabic digits transfers to better math proficiency. The control training was based on discriminating quantity representations, without involving digits. Each of these three-week computer-based trainings were added to the school schedule. We measured improvements in approximate and exact arithmetic after training. Both the experimental and the control group improved in approximate arithmetic performance. However, in exact arithmetic, results show that strengthening the digit-quantity relation improved the 7-year-olds’ competence in symbolic additions and subtractions over and above the improvement measured in the control group. Our results speak to the complexity of the factors involved in developing mathematical abilities, making the case that training the mapping from estimated quantities to digits can be particularly effective in improving children’s mathematical performance.