arXiv Science⌕ Search

SEARCH · arXiv Science

Results for “cs.SC”

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 397 records · Page 22Linked to original sources

Explicit Noether Normalization for Simultaneous Conjugation via Polynomial Identity Testing

Mulmuley recently gave an explicit version of Noether's Normalization lemma for ring of invariants of matrices under simultaneous conjugation, under the conjecture that there are deterministic black-box algorithms for polynomial identity testing (PIT). He argued that this gives evidence that constructing such algorithms for PIT is beyond current techniques. In this work, we show this is not the case. That is, we improve Mulmuley's reduction and correspondingly weaken the conjecture regarding PIT needed to give explicit Noether Normalization. We then observe that the weaker conjecture has recently been nearly settled by the authors, who gave quasipolynomial size hitting sets for the class of read-once oblivious algebraic branching programs (ROABPs). This gives the desired explicit Noether Normalization unconditionally, up to quasipolynomial factors. As a consequence of our proof we give a deterministic parallel polynomial-time algorithm for deciding if two matrix tuples have intersecting orbit closures, under simultaneous conjugation. We also study the strength of conjectures that Mulmuley requires to obtain similar results as ours. We prove that his conjectures are stronger, in the sense that the computational model he needs PIT algorithms for is equivalent to the well-known algebraic branching program (ABP) model, which is provably stronger than the ROABP model. Finally, we consider the depth-3 diagonal circuit model as defined by Saxena, as PIT algorithms for this model also have implications in Mulmuley's work. Previous work have given quasipolynomial size hitting sets for this model. In this work, we give a much simpler construction of such hitting sets, using techniques of Shpilka and Volkovich.

cs.CC↗

Elimination for generic sparse polynomial systems

We present a new probabilistic symbolic algorithm that, given a variety defined in an n-dimensional affine space by a generic sparse system with fixed supports, computes the Zariski closure of its projection to an l-dimensional coordinate affine space with l < n. The complexity of the algorithm depends polynomially on combinatorial invariants associated to the supports.

math.AG↗

New Symbolic Algorithms For Solving A General Bordered Tridiagonal Linear System

In this paper, the author present reliable symbolic algorithms for solving a general bordered tridiagonal linear system. The first algorithm is based on the LU decomposition of the coefficient matrix and the computational cost of it is O(n). The second is based on The Sherman-Morrison-Woodbury formula. The algorithms are implementable to the Computer Algebra System (CAS) such as MAPLE, MATLAB and MATHEMATICA. Three examples are presented for the sake of illustration.

cs.SC↗

Separating linear forms for bivariate systems

We present an algorithm for computing a separating linear form of a system of bivariate polynomials with integer coefficients, that is a linear combination of the variables that takes different values when evaluated at distinct (complex) solutions of the system. In other words, a separating linear form defines a shear of the coordinate system that sends the algebraic system in generic position, in the sense that no two distinct solutions are vertically aligned. The computation of such linear forms is at the core of most algorithms that solve algebraic systems by computing rational parameterizations of the solutions and, moreover, the computation a separating linear form is the bottleneck of these algorithms, in terms of worst-case bit complexity. Given two bivariate polynomials of total degree at most $d$ with integer coefficients of bitsize at most~$τ$, our algorithm computes a separating linear form in $\sOB(d^{8}+d^7τ)$ bit operations in the worst case, where the previously known best bit complexity for this problem was $\sOB(d^{10}+d^9τ)$ (where $\sO$ refers to the complexity where polylogarithmic factors are omitted and $O_B$ refers to the bit complexity).

cs.SC↗

Rational Univariate Representations of Bivariate Systems and Applications

We address the problem of solving systems of two bivariate polynomials of total degree at most $d$ with integer coefficients of maximum bitsize $τ$. It is known that a linear separating form, that is a linear combination of the variables that takes different values at distinct solutions of the system, can be computed in $\sOB(d^{8}+d^7τ)$ bit operations (where $O_B$ refers to bit complexities and $\sO$ to complexities where polylogarithmic factors are omitted) and we focus here on the computation of a Rational Univariate Representation (RUR) given a linear separating form. We present an algorithm for computing a RUR with worst-case bit complexity in $\sOB(d^7+d^6τ)$ and bound the bitsize of its coefficients by $\sO(d^2+dτ)$. We show in addition that isolating boxes of the solutions of the system can be computed from the RUR with $\sOB(d^{8}+d^7τ)$ bit operations. Finally, we show how a RUR can be used to evaluate the sign of a bivariate polynomial (of degree at most $d$ and bitsize at most $τ$) at one real solution of the system in $\sOB(d^{8}+d^7τ)$ bit operations and at all the $Θ(d^2)$ {real} solutions in only $O(d)$ times that for one solution.

cs.SC↗

Numerical method for real root isolation of semi-algebraic system and its applications

In this paper, based on the homotopy continuation method and the interval Newton method, an efficient algorithm is introduced to isolate the real roots of semi-algebraic system. Tests on some random examples and a variety of problems including transcendental functions arising in many applications show that the new algorithm reduces the cost substantially compared with the traditional symbolic approaches.

math.NA↗

Highly Scalable Multiplication for Distributed Sparse Multivariate Polynomials on Many-core Systems

We present a highly scalable algorithm for multiplying sparse multivariate polynomials represented in a distributed format. This algo- rithm targets not only the shared memory multicore computers, but also computers clusters or specialized hardware attached to a host computer, such as graphics processing units or many-core coprocessors. The scal- ability on the large number of cores is ensured by the lacks of synchro- nizations, locks and false-sharing during the main parallel step.

cs.SC↗

Synthesizing Switching Controllers for Hybrid Systems by Continuous Invariant Generation

We extend a template-based approach for synthesizing switching controllers for semi-algebraic hybrid systems, in which all expressions are polynomials. This is achieved by combining a QE (quantifier elimination)-based method for generating continuous invariants with a qualitative approach for predefining templates. Our synthesis method is relatively complete with regard to a given family of predefined templates. Using qualitative analysis, we discuss heuristics to reduce the numbers of parameters appearing in the templates. To avoid too much human interaction in choosing templates as well as the high computational complexity caused by QE, we further investigate applications of the SOS (sum-of-squares) relaxation approach and the template polyhedra approach in continuous invariant generation, which are both well supported by efficient numerical solvers.

eess.SY↗

A probabilistic algorithm to compute the real dimension of a semi-algebraic set

Let $\RR$ be a real closed field (e.g. the field of real numbers) and $\mathscr{S} \subset \RR^n$ be a semi-algebraic set defined as the set of points in $\RR^n$ satisfying a system of $s$ equalities and inequalities of multivariate polynomials in $n$ variables, of degree at most $D$, with coefficients in an ordered ring $\ZZ$ contained in $\RR$. We consider the problem of computing the {\em real dimension}, $d$, of $\mathscr{S}$. The real dimension is the first topological invariant of interest; it measures the number of degrees of freedom available to move in the set. Thus, computing the real dimension is one of the most important and fundamental problems in computational real algebraic geometry. The problem is ${\rm NP}_{\mathbb{R}}$-complete in the Blum-Shub-Smale model of computation. The current algorithms (probabilistic or deterministic) for computing the real dimension have complexity $(s \, D)^{O(d(n-d))}$, that becomes $(s \, D)^{O(n^2)}$ in the worst-case. The existence of a probabilistic or deterministic algorithm for computing the real dimension with single exponential complexity with a factor better than ${O(n^2)}$ in the exponent in the worst-case, is a longstanding open problem. We provide a positive answer to this problem by introducing a probabilistic algorithm for computing the real dimension of a semi-algebraic set with complexity $(s\, D)^{O(n)}$.

cs.SC↗

Symbolic Arithmetic and Integer Factorization

In this paper, we create a systematic and automatic procedure for transforming the integer factorization problem into the problem of solving a system of Boolean equations. Surprisingly, the resulting system of Boolean equations takes on a "life of its own" and becomes a new type of integer, which we call a generic integer. We then proceed to use the newly found algebraic structure of the ring of generic integers to create two new integer factoring algorithms, called respectively the Boolean factoring (BF) algorithm, and the multiplicative Boolean factoring (MBF) algorithm. Although these two algorithms are not competitive with current classical integer factoring algorithms, it is hoped that they will become stepping stones to creating much faster and more competitive algorithms, and perhaps be precursors of a new quantum algorithm for integer factoring.

math.NT↗

Simplifying Multiple Sums in Difference Fields

In this survey article we present difference field algorithms for symbolic summation. Special emphasize is put on new aspects in how the summation problems are rephrased in terms of difference fields, how the problems are solved there, and how the derived results in the given difference field can be reinterpreted as solutions of the input problem. The algorithms are illustrated with the Mathematica package \SigmaP\ by discovering and proving new harmonic number identities extending those from (Paule and Schneider, 2003). In addition, the newly developed package \texttt{EvaluateMultiSums} is introduced that combines the presented tools. In this way, large scale summation problems for the evaluation of Feynman diagrams in QCD (Quantum ChromoDynamics) can be solved completely automatically.

cs.SC↗

Efficient Calculation of Determinants of Symbolic Matrices with Many Variables

Efficient matrix determinant calculations have been studied since the 19th century. Computers expand the range of determinants that are practically calculable to include matrices with symbolic entries. However, the fastest determinant algorithms for numerical matrices are often not the fastest for symbolic matrices with many variables. We compare the performance of two algorithms, fraction-free Gaussian elimination and minor expansion, on symbolic matrices with many variables. We show that, under a simplified theoretical model, minor expansion is faster in most situations. We then propose optimizations for minor expansion and demonstrate their effectiveness with empirical data.

cs.SC↗

Intrinsic complexity estimates in polynomial optimization

It is known that point searching in basic semialgebraic sets and the search for globally minimal points in polynomial optimization tasks can be carried out using $(s\,d)^{O(n)}$ arithmetic operations, where $n$ and $s$ are the numbers of variables and constraints and $d$ is the maximal degree of the polynomials involved.\spar \noindent We associate to each of these problems an intrinsic system degree which becomes in worst case of order $(n\,d)^{O(n)}$ and which measures the intrinsic complexity of the task under consideration.\spar \noindent We design non-uniformly deterministic or uniformly probabilistic algorithms of intrinsic, quasi-polynomial complexity which solve these problems.

cs.SC↗

Plane mixed discriminants and toric jacobians

Polynomial algebra offers a standard approach to handle several problems in geometric modeling. A key tool is the discriminant of a univariate polynomial, or of a well-constrained system of polynomial equations, which expresses the existence of a multiple root. We concentrate on bivariate polynomials and establish an original formula that relates the mixed discriminant of two bivariate Laurent polynomials with fixed support, with the sparse resultant of these polynomials and their toric Jacobian. This allows us to obtain a new proof for the bidegree of the mixed discriminant as well as to establish multipicativity formulas arising when one polynomial can be factored.

math.AG↗

Polynomial Systems Solving by Fast Linear Algebra

Polynomial system solving is a classical problem in mathematics with a wide range of applications. This makes its complexity a fundamental problem in computer science. Depending on the context, solving has different meanings. In order to stick to the most general case, we consider a representation of the solutions from which one can easily recover the exact solutions or a certified approximation of them. Under generic assumption, such a representation is given by the lexicographical Gröbner basis of the system and consists of a set of univariate polynomials. The best known algorithm for computing the lexicographical Gröbner basis is in $\widetilde{O}(d^{3n})$ arithmetic operations where $n$ is the number of variables and $d$ is the maximal degree of the equations in the input system. The notation $\widetilde{O}$ means that we neglect polynomial factors in $n$. We show that this complexity can be decreased to $\widetilde{O}(d^{ωn})$ where $2 \leq ω< 2.3727$ is the exponent in the complexity of multiplying two dense matrices. Consequently, when the input polynomial system is either generic or reaches the Bézout bound, the complexity of solving a polynomial system is decreased from $\widetilde{O}(D^3)$ to $\widetilde{O}(D^ω)$ where $D$ is the number of solutions of the system. To achieve this result we propose new algorithms which rely on fast linear algebra. When the degree of the equations are bounded uniformly by a constant we propose a deterministic algorithm. In the unbounded case we present a Las Vegas algorithm.

cs.SC↗

Harmonic Sums, Polylogarithms, Special Numbers, and their Generalizations

In these introductory lectures we discuss classes of presently known nested sums, associated iterated integrals, and special constants which hierarchically appear in the evaluation of massless and massive Feynman diagrams at higher loops. These quantities are elements of stuffle and shuffle algebras implying algebraic relations being widely independent of the special quantities considered. They are supplemented by structural relations. The generalizations are given in terms of generalized harmonic sums, (generalized) cyclotomic sums, and sums containing in addition binomial and inverse-binomial weights. To all these quantities iterated integrals and special numbers are associated. We also discuss the analytic continuation of nested sums of different kind to complex values of the external summation bound N.

math-ph↗