Fast Estimates of Hankel Matrix Condition Numbers and Numeric Sparse Interpolation*
|
|
- Georgia Jenkins
- 5 years ago
- Views:
Transcription
1 Fast Estimates of Hankel Matrix Condition Numbers and Numeric Sparse Interpolation* Erich L Kaltofen Department of Mathematics North Carolina State University Raleigh, NC , USA kaltofen@mathncsuedu wwwkaltofenus Zhengfeng Yang Shanghai Key Laboratory of Trustworthy Computing East China Normal University Shanghai 0006, China zfyang@seiecnueducn Wen-shin Lee Mathematics and Computer Science Department University of Antwerp 00 Antwerpen, Belgium wen-shinlee@uaacbe ABSTRACT We investigate our early termination criterion for sparse polynomial interpolation when substantial noise is present in the values of the polynomial Our criterion in the exact case uses Monte Carlo randomization which introduces a second source of error We harness the Gohberg-Semencul formula for the inverse of a Hankel matrix to compute estimates for the structured condition numbers of all arising Hankel matrices in quadratic arithmetic time overall, and explain how false ill-conditionedness can arise from our randomizations Finally, we demonstrate by experiments that our condition number estimates lead to a viable termination criterion for polynomials with about 0 non-zero terms and of degree about 00, even in the presence of noise of relative magnitude 0 Categories and Subect Descriptors F [Numerical Algorithms and Problems]: Computations on matrices, Computations on polynomials; G [Interpolation]; I [Algorithms]: Algebraic algorithms, Analysis of algorithms General Terms Algorithm, Theory This material is based on work supported in part by the National Science Foundation under Grant CCF-0807 (Kaltofen) and a 00 Ky and Yu-Fen Fan Fund (Kaltofen and Yang) This material is supported by the Chinese National Natural Science Foundation under Grant 0900 and the Shanghai Natural Science Foundation under Grant 09ZR08800 (Yang) Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee SNC 0, June 7 9, 0, San Jose, California, USA Copyright 0 ACM $000 Keywords Gohberg-Semencul formula, Hankel system, Vandermonde system, sparse polynomial interpolation, early termination INTRODUCTION The fast solution of Toeplitz- and Hankel-like linear systems is based on the classical 97 Levinson-Durbin problem of computing linear generators of the sequence of entries When with exact arithmetic on the entries a singular leading principal submatrix is encountered, look-ahead can be performed: this is the genesis of the 968 Berlekamp/ Massey algorithm In the numeric setting, the notion of ill-conditionedness substitutes for exact singularity, and the algorithms need to estimate the condition numbers of the arising submatrices By a result of Siegfried Rump [], the structured condition number of an n n Toeplitz/Hankel matrix H is equal to its unstructured condition number κ (H): we have a Hankel perturbation matrix H with spectral (matrix-) norm H = / H = H /κ (H) such that H + H is singular [, Theorem ] Thus Trench s [] (exact) inverse algorithm can produce upper and lower bounds for those distances from the Frobenius (Euclideanvector-) norm estimates for / n H F H H F The look-ahead algorithms, such as the Berlekamp/Massey solver, however, need such estimates for all k k leading principal submatrices H [k] for k =,, of an unbounded Hankel matrix Here we show how the 97 Gohberg-Semencul off-diagonal (JL)U-representations [7, 7] can give estimates for all κ (H [k] ) in quadratic arithmetic overall time Although our estimates are quite accurate (see Figures and ), the computation of the actual condition numbers of all leading principal submatrices in quadratic arithmetic time remains an intriguing open problem We note that fraction-free implementations [, ] of the Berlekamp/Massey algorithms and inverse generators only produce values of polynomials in the entries, and our numeric Schwartz/Zippel lemma [0, Lemma ] is applicable In Section we will discuss the numeric sensitivity of the expected values in that Lemma Our investigations are motivated by numeric sparse polynomial interpolation algorithms [,, 0, ] In the Ben-Or/Tiwari version of this algorithm, one first computes 0
2 for a sparse polynomial f(x) C[x] the linear generator for h l = f(ω l+ ) for l = 0,,, In some exact as well as numeric algorithms, ω is selected as a random p-th root of unity Then by the theory of early termination of that algorithm [8], the first t t leading principle submatrix H [t] in the (infinite) Hankel matrix H with entries in row i and column, (H) i, = h i+ where k =,,, t, h 0 h h h h k h h h h h k H [k] = h h h h h k+ h h h h 6 () 6 7 h k h k h k+ h k that is singular has with high probability dimension t = the number of non-zero terms in f For technical reasons in proof of the probabilistic analysis, we must skip over the value f(ω 0 ) = f() Hence, by the previously described algorithm for lower and upper bounds of the condition numbers of H [k], we can determine the number of terms t The Berlekamp/Massey algorithm solves the problem of computing Padé forms along a diagonal of the Padé table Cabay and Meleshko gave an algorithm to bound from above the condition numbers of all the principal submatrices of the associated Hankel system in the well-conditioned case [8], which has been further refined and extended, for example, see [, 6] Here we focus less on stably computing the Gohberg-Semencul updates (see those cited papers we require no look-ahead), but on fast and accurate lower and upper bounds of the condition numbers, which constitute the early termination criterion We add that our approach here, namely testing submatrices for ill-conditionedness, is problematic for the approximate GCD problem (cf [, ]): we know now that Rump s property is false for Sylvester matrices: the unstructured condition number of a Sylvester matrix can be large while the structured condition number is small (see [9, Example ]) We approach the problem of incrementally estimating Hankel submatrices through a numerical interpretation of the Berlekamp/Massey algorithm [] that combines with an application of the numerical Schwartz/Zippel lemma [0] We report some experimental results of our implementation in Section For a 0 term polynomial of degree 00 the algorithm correctly computes t = 0 in quadratic time (given the polynomial evaluations h l ) (see Table ) There are several issues to scrutinize First, we only have bounds on the condition numbers and must verify that our estimates are sufficiently accurate One would expect since for h l = f(ω l+ ) the Hankel matrix H [k] in () has additional structure that the upper bound estimates for the Rump condition numbers are good indicators of numeric non-singularity Second, early termination is achieved by randomization, whose probabilistic analysis, namely the separation of the decision quantities, ie, determinants or condition numbers, from very small values (in terms of absolute values), applies to exact computation We can relate the numeric sensitivity of those decision quantities via a factorization of H [t] to clustering of term values on the unite circle, which is partially overcome by evaluation at several random ω simultaneously Throughout the paper, we will use the boldfaced letters x and y to denote the vectors [x,, x n] T, [y,, y n] T, respectively; H [k] denotes the k k leading principle submatrix; κ(h) = H H denotes the unstructured condition number of H CONDITION NUMBER ESTIMATE Suppose a n n Hankel matrix H is strongly regular, ie, all the leading principle submatrices of H are nonsingular Following an efficient algorithm for solving Hankel systems [], we present a recursive method to estimate the condition numbers of all leading principal Hankel submatrices in quadratic time For simplicity, in this section we only consider the condition number of the Hankel matrix with respect to norm, ie, κ (H) = H H, since all other operator norms can be bounded by the corresponding norm In order to estimate the condition numbers of all leading principal submatrices, we need to estimate the norms of all leading principal submatrices as well as their inverses Since a Hankel matrix H has recursive structure, the norm of H [i+] can be obtained in O(i) flops if the norm of H [i] is given, which implies that computing the norms of all leading principal submatrices can be done in O(n ) flops Therefore, our task is to demonstrate how to estimate the norm of (H [i+] ) in linear time if one has the estimate of the norm of (H [i] ) As restated in Theorem, the Gohberg-Semencul formula for inverting a Hankel matrix plays an important role in our method Note that for a given Hankel matrix H, J H is a Toeplitz matrix, where J is an anti-diagonal matrix with as its nonzero entries Theorem Given a Hankel matrix H, suppose x and y are the solution vectors of the systems of the equations Hx = e, Hy = e n, where e = [, 0, 0,, 0] T and e n = [0,, 0, ] T Then if x n 0 the inverse H satisfies the identity H = x x x n x n y y y n y n x x x n 0 0 y y n y n x n x n x n y y x n } y } JL R y y y n 0 0 x x n x n y y x n x n x n y n x () } } JL R The inversion formula of a Hankel matrix leads to a polynomial form, whose coefficients involve vectors x and y Theorem Under the assumptions of Theorem, suppose H = [γ, γ,, γ n], where γ k denotes the k-th column of H Then we have γ k = [v n, v n,, v 0] T, where v i denotes the coefficient of the polynomial (/x n)
3 (φ g k u k v) with respect to Z i, where φ = v = nx x iz n i, g k = i= nx i= kx y iz k i, i= k X y iz n i, u = 0, u k = x iz k i, k =,, n Proof Considering the polynomial product φ g k, we obtain the coefficient subvector i= µ = [ψ n, ψ n,, ψ 0] T, where ψ is the coefficient of φ g k with respect to Z Expanding φ g k, we get P ψ = i=0 xn iy k+i, if 0 k, P k i=0 xn+i y k i, if k < < n Observing the formula of H in Theorem, We find that the coefficient subvector µ = JL [y k, y k,, 0, 0] T, () where [y k, y k,, 0, 0] is the k-th column of R Since u = 0, it is obvious that the coefficient vector of u v is a zero vector Considering the polynomial product u k v, k n, we obtain the coefficient subvector ν = [χ n, χ n,, χ 0] T, where χ is the coefficient of u k v with respect to Z Expanding u k v, we get 8 < 0 if = 0 P χ = i=0 : yn ix k +i, if k, P k i= yn+ x k i, if k < n The coefficient subvector ν = JL [x k, x k,, 0, 0] T, () where [x k, x k,, 0, 0] T is the k-th column of R According to () and (), it can be verified that the column γ k is able to be represented by the coefficients of (/x n) (φ g k u k v) Corollary Given a nonsingular Hankel matrix H, suppose x and y are the solutions of the equations of Hx = e and Hy = e n, then x H x y () x n Proof Since x n = y, () can be concluded from Theorem Given a strongly regular matrix H, we discuss how to use the bound in () to estimate (H [i] ) for all the leading principal submatrices in quadratic time A linear time recursive algorithm is presented in [] to obtain the solution vector of H [i+] x i+ = e i+ from the solution vector of H [i] y i = e i, where H [i] and H [i+] are the i i and (i + ) (i + ) leading principal submatrices of H respectively, and e i = [0,, 0, ] T, e i+ = ˆ0, e i T Moreover, according to the Lemma in [], for an i i strongly regular Hankel matrix H, the solution x of the linear system Hx = e can be obtained in O(i) flops if one has the solution Hy = e i In other words, for a given strongly regular k k Hankel matrix H, all of the solution vectors x and y of H [i] x = e, H [i] y = e i, i =,,, k can be obtained in O(k ) flops Hence the recursive algorithm presented in [] is applicable to obtain the bounds () of all (H [i] ) in quadratic time Theorem Under the same assumptions as above, given a strongly regular n n Hankel matrix H, we have x [i] H [i] κ (H [i] ) x[i] y [i] x [i] i H [i], (6) where x [i] and y [i] are the solution vectors of H [i] x [i] = e and H [i] y [i] = e i, and x [i] i denotes the last element of x [i] Furthermore, computing the lower bounds and the upper bounds (6) of all κ (H [i] ), i n, can be achieved in O(n ) arithmetic operations Proof We conclude (6) from Corollary As discussed, H [i] and the bounds of (H [i] ) can be obtained in quadratic time, which means that the computation of the condition number bounds (6) is O(n ) EARLY TERMINATION IN NUMERICAL SPARSE INTERPOLATION We apply our method to the problem of early termination in numerical sparse polynomial interpolation Consider a black box univariate polynomial f C[x] represented as f(x) = tx c x d, 0 c C (7) = and d Z 0, d < d < < d t Suppose the evaluations of f(x) contain added noise Let δ be a degree upper bound of f, which means, δ deg(f) = d t Our aim is to determine the number of terms t in the target polynomial f In exact arithmetic, the early termination strategy [8] can determine the number of terms with high probability Choose a random element ω from a sufficiently large finite set and evaluate f(x) at the powers of ω h 0 = f(ω), h = f(ω ), h = f(ω ), Consider a (δ + ) (δ + ) Hankel matrix H = (H) i, = h i+ as in () The early termination algorithm makes, with high probability, all principal minors H [] non-singular for t [8] Moreover, H [] must be singular for all t < δ Hence the number of terms t is detected as follows: for =,,, the first time H [] becomes singular is when = t + In the numerical setting, such a singular matrix is often extremely ill-conditioned So we need to measure the singularity for a given numerical Hankel matrix In [], it is proved that for a Hankel matrix the structured condition number with respect to the spectral norm is equivalent to the regular condition number In other words, for a given Hankel matrix H [], the distance measured by spectral norm to the nearest singular Hankel matrix is equivalent to / (H [] ) In the following we investigate how to estimate the spectral norm of the inverse of a Hankel matrix
4 Theorem Using the same quantities as above, we have x n H x y (8) x n Proof Given a n n arbitrary matrix A, we have the following bound of A [6] max{ A, A } n A p A A According to Corollary and H = H, it can be verified that H satisfies (8) We following the bound of H in (8) and present a recursive algorithm to detect the sparsity of the interpolating polynomial Suppose we are given an approximate black box polynomial of f represented as (7), its degree bound δ and a chosen tolerance τ, our aim is to recover the number of terms t We proceed our recursive algorithm as the following First choose a random root of unity ω with a prime order p δ and obtain the approximate evaluations h l f(ω l+ ), l = 0,,, Then we compute iteratively for k =,, and H [k] = [h i+ ] the vectors x [k] and y [k] as stated in [], where x [k] and y [k] are the solutions of H [k] x [k] = e and H [k] y [k] = e k We break out of the loop until x [k] k x [k] y [k] τ, which means that for all =,,, k the distance between H [] and the nearest singular Hankel matrix is greater than the given tolerance τ This is because x [k] k > τ (H [k] ) x [k] y [k] At this stage, we claim that H [k ] is strongly regular and obtain t = k as the number of the terms in f In sparse polynomial interpolation, the associated Hankel matrix H [t] has additional structure that allows a factorization h 0 h h H [] h h h = 6 7 h h h c b 0 0 = b b b t 0 c b b b b t 0 0 c tb t } } V [] D b b b b 6 7, (9) b t b t } (V [] ) T in which b k = ω d k for k t When = t, the condition of the associated Hankel system is linked to the embedded Vandermonde system [, ] (V [t] ) max c (H [t] ) (V [t] ) P t c (0) (see Proposition in []) Since all b are on the unit circle, according to [9] the inverse of the Vandermonde matrix V can be bounded by (V [t] ) max ty t =, k + b b b k t = max t Q k b b k () As for the lower bound, (V [t] ) = achieves the optimal condition if b are evenly distributed on the unit circle (see Example 6 in [0]) In [, ], the sparsity t is given as input So the randomization is used to improve the condition of the overall sparse interpolation problem, where the recovery of both the nonzero terms and the corresponding coefficients depend on the condition of the embedded Vandermonde system While [, ] further exploit a generalized eigenvalue reformulation [] in sparse interpolation, interestingly the condition of the associated generalized eigenvalue problem is still dependent on the condition of the embedded Vandermonde system [, Theorem ] (see [] for further analysis and discussion) Now our purpose now to detect the number of terms In other words, we intend to use the estimated conditions to determine the number of terms t (After t is determined, for polynomial interpolation one still needs to recover the t exponents d, d t and the associated coefficients c,, c t in f) Recall that in exact arithmetic, H [] is singular for > t In a numerical setting, such H [] is expected to be extremely ill-conditioned If H [t] is relatively well-conditioned, then one can expect a surge in the condition number for H [t+] We apply the randomization idea in [, ] to obtain the heuristic of achieving relatively well-conditioned H [t] with high probability Following (), the distribution of b,, b t on the unit circle affects the upper bound on the condition of the Vandermonde system V [t] Let b = e πid /p, b k = e πidk/p for a prime p δ Both b and b k are on the unit circle Let k = d d k, the distance between b and b k is q b b k = ( cos(π k /p)) + sin (π k /p) = p cos(π k /p) For a fixed t, an upper bound of (V [t] ) is determined by Y ««π k min cos, () t p k in which p is a chosen prime order for the primitive roots of unity According to (), the corresponding Hankel system H [t] = V [t] D(V [t] ) T can be better conditioned if p is smaller; and/or
5 the most (or all) of k do not belong to smaller values In order to achieve a better condition, choosing a larger p would require larger k and may require a higher precision Thus we prefer a smaller p Moreover, in general a smaller p does not affect too much on the overall random distribution of b on the unit circle (see Example in Section ) Remark: In the exact case, it may still happen that H [] is singular for t, which corresponds to an ill-conditioned H [] for t in a finite precision environment Therefore we perform our algorithm ζ times in the numerical setting Given ζ Z >0, we choose different random roots of unity ω, ω,, ω ζ For each we perform our algorithm on f(ω k ) for k =,, and obtain the number of terms as t At the end, we determine t = max{t, t,, t ζ } as the number of the terms of f EXPERIMENTS Our numerical early termination is tested for determining the number of the terms We set Digits := in Maple Our test polynomials are constructed with random integer coefficients in the range between 0 and 0, and with a random degree and random terms in the given ranges For each given noise range, we test 0 random black box polynomials and report the times of failing to correctly estimate the number of the terms In Table, Deg Range and Term Range record the range of the degree and the number of the term in the polynomials; Random Noise records range of random noise added to the evaluations of the black box polynomial; F ail reports the times of failing to estimate the correct number of the terms in the target polynomials Ex Random Noise Term Range Deg Range F ail Table : Numerical early termination: number of failures out of 0 random polynomials Example Given a polynomial f with deg(f) = 00 and the number of term 0, denoted by T (f) = 0 Its coefficients are randomly generated as integers between 0 and 0 We construct the black box that evaluates f with added random noises in the range From such black box evaluations, we generated 000 random Hankel matrices and compare the condition number κ (H) = H H with their corresponding lower bounds and the upper bound κ up(h), shown in (6) Let ω = exp(s πi /p ) C be random roots of unity, where i =, p prime numbers in the range 00 p 000, and s random integers with s < p We compute the evaluations of the black box for different root of unity ω, and construct the associated Hankel matrix H [] Since T (f) = 0, we do evaluations for each to construct the Hankel matrix with Dim(H [] ) = : h,l f(ω l ) C, l =,,,, We use H [k] of H and let to denote the k k leading principle submatrix r low = H H κ up(h), r up = H H For =,,, 000, we obtain the ratios r low and r up for all k k leading principle submatrices H [k], where k =,,, 0 Therefore, r low and r up are used to measure whether the lower and upper bounds are close to the actual condition number We use the histogram to measure the ratios r low and r up For instance, in Figure (a) and (b) show the distribution of the ratios r low and r up for all leading principle submatrices H [], =,,, 000 In Figure, (a) and (b) show the distribution of the ratios r low and r up of all 0 0 leading principle submatrices H [0], =,,, 000 Example We performed m = 00 random n n Hankel matrices for n =, 8,, 0 whose entries are randomly chosen and follow uniformly probabilistic distribution over {c c } Table displays the average value of r low and r up Here n is the dimension of random Hankel matrices; r low and r up are the average ratios of H H κ and up(h) ; κ H H = H H is the average value of the actual condition number obtained for Digits := in Maple From the columns of r low and r up, we observe that the upper bound of the condition number depends on the dimension of the Hankel matrices But the lower bound of the condition number does not seem to be affected much by the increasing of the dimension Ex n r low r up κ (H) Table : Algorithm performance on condition number estimate Finally, following the discussion on (0), (), () in Section, we investigate the role of p in the conditioning of H [t] Example For each try, we choose m = 0 random polynomials with 0 deg(f) 00 The coefficients are randomly generated as integers between and The number of the terms of f is t, shown in the -th column of Table In Table, Term Range records the range of random noise added to the evaluations of the black box polynomial; κ (H [t] ) and κ (H [t] ) denote the respective condition numbers of the t t Hankel matrices H [t] and H [t] The matrices H and H are constructed by the approximate [t] [t] evaluations of the root of unity with orders corresponding to
6 (a) Distribution r low = H H κ up(h) (b) Distribution r up = H H Figure : Distribution of the leading principle matrix H [] (a) Distribution r low = H H κ up(h) (b) Distribution r up = H H Figure : Distribution of the leading principle matrix H [0] the randomly chosen prime numbers p and p, where p is between 00 and 000 and p between 0 and 0 7 Table agrees with our discussion at the end of Section When the number of terms t is fixed, the various scales of p and p do not seem to greatly affect the distribution of terms on the unit circle But a larger p may demand a higher precision hence cause additional computation problems CONCLUSION By example of a randomized sparse interpolation algorithm, we have investigated a central question in algorithms with floating point arithmetic and imprecise date posed in [0]: how does one analyze the probability of success, in our case the correct determination of the discrete polynomial sparsity? Bad random choices may result in ill-conditioned intermediate structured matrices at the wrong place Even if one can detect such ill-conditionedness, as we have via the Gohberg-Semencul formula, the distribution of such occurrences must be controlled We do so by running multiple random execution paths next to one another, which successfully can weed out bad random choices, as we have observed experimentally The mathematical ustification requires an understanding of condition numbers of random Fourier matrices, which we have presented, and of additional instabilities that are caused by substantial noise in the data, which we have demonstrated, at least by experiment, to be con-
7 Ex Random Noise t κ (H [t] ) κ (H [t] ) Table : Condition numbers for H [t] constructed with roots of unity of different size prime orders p and p trollable by our algorithms Acknowledgments We thank anonymous referees for their helpful comments and information 6 REFERENCES [] B Beckermann The stable computation of formal orthogonal polynomials Numerical Algorithms, ():, 996 [] B Beckermann, S Cabay, and G Labahn Fraction-free computation of matrix Pade systems In W Küchlin, editor, Proc 997 Internat Symp Symbolic Algebraic Comput (ISSAC 97), pages, 997 [] B Beckermann, G H Golub, and G Labahn On the numerical condition of a generalized Hankel eigenvalue problem Numerische Mathematik, 06(): 68, 007 [] B Beckermann and G Labahn A fast and numerically stable euclidean-like algorithm for detecting relatively prime numerical polynomials Journal of Symbolic Computation, 6(6):69 7, 998 [] B Beckermann and G Labahn When are two numerical polynomials relatively prime? Journal of Symbolic Computation, 6(6): , 998 [6] B Beckermann and G Labahn Effective computation of rational approximants and interpolants Reliable Computing, 6():6 90, 000 [7] R P Brent, F G Gustavson, and D Y Y Yun Fast solution of Toeplitz systems of equations and computation of Padé approximants J Algorithms, :9 9, 980 [8] S Cabay and R Meleshko A weakly stable algorithm for Padé approximants and the inversion of Hankel matrices SIAM Journal on Matrix Analysis and Applicatins, ():7 78, 99 [9] W Gautschi On inverse of Vandermonde and confluent Vandermonde matrices Numerische Mathematik, :7, 96 [0] W Gautschi Norm estimates for inverse of Vandermonde matrices Numerische Mathematik, :7 7, 97 [] M Giesbrecht, G Labahn, and W Lee Symbolic-numeric sparse interpolation of multivariate polynomials (extended abstract) In J-G Dumas, editor, ISSAC MMVI Proc 006 Internat Symp Symbolic Algebraic Comput, pages 6 ACM Press, 006 [] M Giesbrecht, G Labahn, and W Lee Symbolic-numeric sparse interpolation of multivariate polynomials Journal of Symbolic Computation, :9 99, 009 [] M Giesbrecht and D Roche Diversification improves interpolation In A Leykin, editor, ISSAC 0 Proc 0 Internat Symp Symbolic Algebraic Comput ACM Press, 0 to appear [] I Gohberg and I Koltracht Efficient algorithm for Toeplitz plus Hankel matrices Integral Equations and Operator Theory, :6, 989 [] G H Golub, P Milanfar, and J Varah A stable numerical method for inverting shape from moments SIAM Journal on Scientific Computing, ():, 999 [6] G H Golub and C F Van Loan Matrix Computations Johns Hopkins University Press, Baltimore, Maryland, third edition, 996 [7] T Huckle Computations with Gohberg-Semencul formulas for Toeplitz matrices Linear Algebra and Applications, 7:69 98, 998 [8] E Kaltofen and W Lee Early termination in sparse interpolation algorithms Journal of Symbolic Computation, 6( ):6 00, 00 URL: wwwmathncsuedu/ kaltofen/bibliography/0/kl0 pdf [9] E Kaltofen, Z Yang, and L Zhi Approximate greatest common divisors of several polynomials with linearly constrained coefficients and singular polynomials In J-G Dumas, editor, ISSAC MMVI Proc 006 Internat Symp Symbolic Algebraic Comput, pages 69 76, New York, N Y, 006 ACM Press URL: kaltofen/ bibliography/06/kyz06pdf [0] E Kaltofen, Z Yang, and L Zhi On probabilistic analysis of randomization in hybrid symbolic-numeric algorithms In S M Watt and J Verschelde, editors, SNC 07 Proc 007 Internat Workshop on Symbolic-Numeric Comput, pages 7, 007 URL: kaltofen/bibliography/ 07/KYZ07pdf [] E Kaltofen and G Yuhasz A fraction free matrix Berlekamp/Massey algorithm, Feb 009 Manuscript, 7 pages [] S M Rump Structured perturbations part I: Normwise distances SIAM Journal on Matrix Analysis and Applications, (): 0, 00 [] W F Trench An algorithm for the inversion of finite Hankel matrices Journal of the Society for Industrial and Applied Mathematics, ():pp 0 07, 96 6
Journal of Symbolic Computation. On the Berlekamp/Massey algorithm and counting singular Hankel matrices over a finite field
Journal of Symbolic Computation 47 (2012) 480 491 Contents lists available at SciVerse ScienceDirect Journal of Symbolic Computation journal homepage: wwwelseviercom/locate/jsc On the Berlekamp/Massey
More informationOn the Berlekamp/Massey Algorithm and Counting Singular Hankel Matrices over a Finite Field
On the Berlekamp/Massey Algorithm and Counting Singular Hankel Matrices over a Finite Field Matthew T Comer Dept of Mathematics, North Carolina State University Raleigh, North Carolina, 27695-8205 USA
More informationComputing Approximate GCD of Univariate Polynomials by Structure Total Least Norm 1)
MM Research Preprints, 375 387 MMRC, AMSS, Academia Sinica No. 24, December 2004 375 Computing Approximate GCD of Univariate Polynomials by Structure Total Least Norm 1) Lihong Zhi and Zhengfeng Yang Key
More informationA Fraction Free Matrix Berlekamp/Massey Algorithm*
Linear Algebra and Applications, 439(9):2515-2526, November 2013. A Fraction Free Matrix Berlekamp/Massey Algorithm* Erich Kaltofen and George Yuhasz Dept. of Mathematics, North Carolina State University,
More informationThe Seven Dwarfs of Symbolic Computation and the Discovery of Reduced Symbolic Models. Erich Kaltofen North Carolina State University google->kaltofen
The Seven Dwarfs of Symbolic Computation and the Discovery of Reduced Symbolic Models Erich Kaltofen North Carolina State University google->kaltofen Outline 2 The 7 Dwarfs of Symbolic Computation Discovery
More informationAn Algorithm for Approximate Factorization of Bivariate Polynomials 1)
MM Research Preprints, 402 408 MMRC, AMSS, Academia Sinica No. 22, December 2003 An Algorithm for Approximate Factorization of Bivariate Polynomials 1) Zhengfeng Yang and Lihong Zhi 2) Abstract. In this
More informationInteger Least Squares: Sphere Decoding and the LLL Algorithm
Integer Least Squares: Sphere Decoding and the LLL Algorithm Sanzheng Qiao Department of Computing and Software McMaster University 28 Main St. West Hamilton Ontario L8S 4L7 Canada. ABSTRACT This paper
More informationBlack Box Linear Algebra
Black Box Linear Algebra An Introduction to Wiedemann s Approach William J. Turner Department of Mathematics & Computer Science Wabash College Symbolic Computation Sometimes called Computer Algebra Symbols
More informationFast algorithms for factoring polynomials A selection. Erich Kaltofen North Carolina State University google->kaltofen google->han lu
Fast algorithms for factoring polynomials A selection Erich Kaltofen North Carolina State University google->kaltofen google->han lu Overview of my work 1 Theorem. Factorization in K[x] is undecidable
More informationOn Exact and Approximate Interpolation of Sparse Rational Functions*
On Exact and Approximate Interpolation of Sparse Rational Functions* ABSTRACT Erich Kaltofen Department of Mathematics North Carolina State University Raleigh, North Carolina 27695-8205, USA kaltofen@math.ncsu.edu
More informationSparse Polynomial Interpolation and Berlekamp/Massey algorithms that correct Outlier Errors in Input Values
Sparse Polynomial Interpolation and Berlekamp/Massey algorithms that correct Outlier Errors in Input Values Clément PERNET joint work with Matthew T. COMER and Erich L. KALTOFEN : LIG/INRIA-MOAIS, Grenoble
More informationEigen-solving via reduction to DPR1 matrices
Computers and Mathematics with Applications 56 (2008) 166 171 www.elsevier.com/locate/camwa Eigen-solving via reduction to DPR1 matrices V.Y. Pan a,, B. Murphy a, R.E. Rosholt a, Y. Tang b, X. Wang c,
More informationAlgorithm 853: an Efficient Algorithm for Solving Rank-Deficient Least Squares Problems
Algorithm 853: an Efficient Algorithm for Solving Rank-Deficient Least Squares Problems LESLIE FOSTER and RAJESH KOMMU San Jose State University Existing routines, such as xgelsy or xgelsd in LAPACK, for
More informationABSTRACT. YUHASZ, GEORGE L. Berlekamp/Massey Algorithms for Linearly Generated Matrix Sequences. (Under the direction of Professor Erich Kaltofen.
ABSTRACT YUHASZ, GEORGE L. Berlekamp/Massey Algorithms for Linearly Generated Matrix Sequences. (Under the direction of Professor Erich Kaltofen.) The Berlekamp/Massey algorithm computes the unique minimal
More informationDeterministic distinct-degree factorisation of polynomials over finite fields*
Article Submitted to Journal of Symbolic Computation Deterministic distinct-degree factorisation of polynomials over finite fields* Shuhong Gao 1, Erich Kaltofen 2 and Alan G.B. Lauder 3 1 Department of
More informationWe first repeat some well known facts about condition numbers for normwise and componentwise perturbations. Consider the matrix
BIT 39(1), pp. 143 151, 1999 ILL-CONDITIONEDNESS NEEDS NOT BE COMPONENTWISE NEAR TO ILL-POSEDNESS FOR LEAST SQUARES PROBLEMS SIEGFRIED M. RUMP Abstract. The condition number of a problem measures the sensitivity
More informationDisplacement Structure in Computing Approximate GCD of Univariate Polynomials
MM Research Preprints, 230 238 MMRC, AMSS, Academia, Sinica, Beijing No 21, December 2002 Displacement Structure in Computing Approximate GCD of Univariate Polynomials Lihong Zhi 1) Abstract We propose
More informationAdaptive decoding for dense and sparse evaluation/interpolation codes
Adaptive decoding for dense and sparse evaluation/interpolation codes Clément PERNET INRIA/LIG-MOAIS, Grenoble Université joint work with M. Comer, E. Kaltofen, J-L. Roch, T. Roche Séminaire Calcul formel
More informationM. VAN BAREL Department of Computing Science, K.U.Leuven, Celestijnenlaan 200A, B-3001 Heverlee, Belgium
MATRIX RATIONAL INTERPOLATION WITH POLES AS INTERPOLATION POINTS M. VAN BAREL Department of Computing Science, K.U.Leuven, Celestijnenlaan 200A, B-3001 Heverlee, Belgium B. BECKERMANN Institut für Angewandte
More informationSection III.6. Factorization in Polynomial Rings
III.6. Factorization in Polynomial Rings 1 Section III.6. Factorization in Polynomial Rings Note. We push several of the results in Section III.3 (such as divisibility, irreducibility, and unique factorization)
More informationFraction-free Row Reduction of Matrices of Skew Polynomials
Fraction-free Row Reduction of Matrices of Skew Polynomials Bernhard Beckermann Laboratoire d Analyse Numérique et d Optimisation Université des Sciences et Technologies de Lille France bbecker@ano.univ-lille1.fr
More informationVector Rational Number Reconstruction
Vector Rational Number Reconstruction Curtis Bright cbright@uwaterloo.ca David R. Cheriton School of Computer Science University of Waterloo, Ontario, Canada N2L 3G1 Arne Storjohann astorjoh@uwaterloo.ca
More informationRITZ VALUE BOUNDS THAT EXPLOIT QUASI-SPARSITY
RITZ VALUE BOUNDS THAT EXPLOIT QUASI-SPARSITY ILSE C.F. IPSEN Abstract. Absolute and relative perturbation bounds for Ritz values of complex square matrices are presented. The bounds exploit quasi-sparsity
More informationSome results on the existence of t-all-or-nothing transforms over arbitrary alphabets
Some results on the existence of t-all-or-nothing transforms over arbitrary alphabets Navid Nasr Esfahani, Ian Goldberg and Douglas R. Stinson David R. Cheriton School of Computer Science University of
More informationSolving Sparse Rational Linear Systems. Pascal Giorgi. University of Waterloo (Canada) / University of Perpignan (France) joint work with
Solving Sparse Rational Linear Systems Pascal Giorgi University of Waterloo (Canada) / University of Perpignan (France) joint work with A. Storjohann, M. Giesbrecht (University of Waterloo), W. Eberly
More informationChapter 3 Transformations
Chapter 3 Transformations An Introduction to Optimization Spring, 2014 Wei-Ta Chu 1 Linear Transformations A function is called a linear transformation if 1. for every and 2. for every If we fix the bases
More informationOn The Belonging Of A Perturbed Vector To A Subspace From A Numerical View Point
Applied Mathematics E-Notes, 7(007), 65-70 c ISSN 1607-510 Available free at mirror sites of http://www.math.nthu.edu.tw/ amen/ On The Belonging Of A Perturbed Vector To A Subspace From A Numerical View
More informationComputing Minimal Nullspace Bases
Computing Minimal Nullspace ases Wei Zhou, George Labahn, and Arne Storjohann Cheriton School of Computer Science University of Waterloo, Waterloo, Ontario, Canada {w2zhou,glabahn,astorjoh}@uwaterloo.ca
More informationComputing with polynomials: Hensel constructions
Course Polynomials: Their Power and How to Use Them, JASS 07 Computing with polynomials: Hensel constructions Lukas Bulwahn March 28, 2007 Abstract To solve GCD calculations and factorization of polynomials
More informationCholesky factorisation of linear systems coming from finite difference approximations of singularly perturbed problems
Cholesky factorisation of linear systems coming from finite difference approximations of singularly perturbed problems Thái Anh Nhan and Niall Madden Abstract We consider the solution of large linear systems
More informationChinese Remainder Theorem
Chinese Remainder Theorem Theorem Let R be a Euclidean domain with m 1, m 2,..., m k R. If gcd(m i, m j ) = 1 for 1 i < j k then m = m 1 m 2 m k = lcm(m 1, m 2,..., m k ) and R/m = R/m 1 R/m 2 R/m k ;
More informationIntroduction. Chapter One
Chapter One Introduction The aim of this book is to describe and explain the beautiful mathematical relationships between matrices, moments, orthogonal polynomials, quadrature rules and the Lanczos and
More informationOPTIMAL SCALING FOR P -NORMS AND COMPONENTWISE DISTANCE TO SINGULARITY
published in IMA Journal of Numerical Analysis (IMAJNA), Vol. 23, 1-9, 23. OPTIMAL SCALING FOR P -NORMS AND COMPONENTWISE DISTANCE TO SINGULARITY SIEGFRIED M. RUMP Abstract. In this note we give lower
More informationA fast randomized algorithm for overdetermined linear least-squares regression
A fast randomized algorithm for overdetermined linear least-squares regression Vladimir Rokhlin and Mark Tygert Technical Report YALEU/DCS/TR-1403 April 28, 2008 Abstract We introduce a randomized algorithm
More informationGeometric interpretation of signals: background
Geometric interpretation of signals: background David G. Messerschmitt Electrical Engineering and Computer Sciences University of California at Berkeley Technical Report No. UCB/EECS-006-9 http://www.eecs.berkeley.edu/pubs/techrpts/006/eecs-006-9.html
More informationNumerical Methods in Matrix Computations
Ake Bjorck Numerical Methods in Matrix Computations Springer Contents 1 Direct Methods for Linear Systems 1 1.1 Elements of Matrix Theory 1 1.1.1 Matrix Algebra 2 1.1.2 Vector Spaces 6 1.1.3 Submatrices
More informationAlgorithms for the Non-monic Case of the Sparse Modular GCD Algorithm
Algorithms for the Non-monic Case of the Sparse Modular GCD Algorithm Jennifer de Kleine Department of Mathematics Simon Fraser University Burnaby, B.C. Canada. dekleine@cecm.sfu.ca. Michael Monagan Department
More informationANONSINGULAR tridiagonal linear system of the form
Generalized Diagonal Pivoting Methods for Tridiagonal Systems without Interchanges Jennifer B. Erway, Roummel F. Marcia, and Joseph A. Tyson Abstract It has been shown that a nonsingular symmetric tridiagonal
More informationToeplitz matrices. Niranjan U N. May 12, NITK, Surathkal. Definition Toeplitz theory Computational aspects References
Toeplitz matrices Niranjan U N NITK, Surathkal May 12, 2010 Niranjan U N (NITK, Surathkal) Linear Algebra May 12, 2010 1 / 15 1 Definition Toeplitz matrix Circulant matrix 2 Toeplitz theory Boundedness
More informationAn algebraic perspective on integer sparse recovery
An algebraic perspective on integer sparse recovery Lenny Fukshansky Claremont McKenna College (joint work with Deanna Needell and Benny Sudakov) Combinatorics Seminar USC October 31, 2018 From Wikipedia:
More informationPrimitive sets in a lattice
Primitive sets in a lattice Spyros. S. Magliveras Department of Mathematical Sciences Florida Atlantic University Boca Raton, FL 33431, U.S.A spyros@fau.unl.edu Tran van Trung Institute for Experimental
More informationBlack Box Linear Algebra with the LinBox Library. William J. Turner North Carolina State University wjturner
Black Box Linear Algebra with the LinBox Library William J. Turner North Carolina State University www.math.ncsu.edu/ wjturner Overview 1. Black Box Linear Algebra 2. Project LinBox 3. LinBox Objects 4.
More informationExact Arithmetic on a Computer
Exact Arithmetic on a Computer Symbolic Computation and Computer Algebra William J. Turner Department of Mathematics & Computer Science Wabash College Crawfordsville, IN 47933 Tuesday 21 September 2010
More informationMETRIC HEIGHTS ON AN ABELIAN GROUP
ROCKY MOUNTAIN JOURNAL OF MATHEMATICS Volume 44, Number 6, 2014 METRIC HEIGHTS ON AN ABELIAN GROUP CHARLES L. SAMUELS ABSTRACT. Suppose mα) denotes the Mahler measure of the non-zero algebraic number α.
More informationFor δa E, this motivates the definition of the Bauer-Skeel condition number ([2], [3], [14], [15])
LAA 278, pp.2-32, 998 STRUCTURED PERTURBATIONS AND SYMMETRIC MATRICES SIEGFRIED M. RUMP Abstract. For a given n by n matrix the ratio between the componentwise distance to the nearest singular matrix and
More informationHighly resilient correctors for polynomials
Highly resilient correctors for polynomials Peter Gemmell Madhu Sudan Abstract We consider the problem of correcting programs that compute multivariate polynomials over large finite fields and give an
More informationReconstruction of sparse Legendre and Gegenbauer expansions
Reconstruction of sparse Legendre and Gegenbauer expansions Daniel Potts Manfred Tasche We present a new deterministic algorithm for the reconstruction of sparse Legendre expansions from a small number
More informationarxiv: v3 [math.ac] 11 May 2016
GPGCD: An iterative method for calculating approximate GCD of univariate polynomials arxiv:12070630v3 [mathac] 11 May 2016 Abstract Akira Terui Faculty of Pure and Applied Sciences University of Tsukuba
More informationResearch Article Minor Prime Factorization for n-d Polynomial Matrices over Arbitrary Coefficient Field
Complexity, Article ID 6235649, 9 pages https://doi.org/10.1155/2018/6235649 Research Article Minor Prime Factorization for n-d Polynomial Matrices over Arbitrary Coefficient Field Jinwang Liu, Dongmei
More informationPreconditioners for Singular Black Box Matrices
Preconditioners for Singular Black Box Matrices William J Turner Department of Mathematics & Computer Science Wabash College Crawfordsville, IN 47933 USA turnerw@wabashedu ABSTRACT Categories and Subject
More informationBlack Box Linear Algebra with the LinBox Library. William J. Turner North Carolina State University wjturner
Black Box Linear Algebra with the LinBox Library William J. Turner North Carolina State University www.math.ncsu.edu/ wjturner Overview 1. Black Box Linear Algebra 2. Project LinBox 3. LinBox Objects 4.
More information18. Cyclotomic polynomials II
18. Cyclotomic polynomials II 18.1 Cyclotomic polynomials over Z 18.2 Worked examples Now that we have Gauss lemma in hand we can look at cyclotomic polynomials again, not as polynomials with coefficients
More informationCS 4770: Cryptography. CS 6750: Cryptography and Communication Security. Alina Oprea Associate Professor, CCIS Northeastern University
CS 4770: Cryptography CS 6750: Cryptography and Communication Security Alina Oprea Associate Professor, CCIS Northeastern University March 15 2018 Review Hash functions Collision resistance Merkle-Damgaard
More informationTHE VECTOR RATIONAL FUNCTION RECONSTRUCTION PROBLEM
137 TH VCTOR RATIONAL FUNCTION RCONSTRUCTION PROBLM ZACH OLSH and ARN STORJOHANN David R. Cheriton School of Computer Science University of Waterloo, Ontario, Canada N2L 3G1 -mail: astorjoh@uwaterloo.ca
More information1 The linear algebra of linear programs (March 15 and 22, 2015)
1 The linear algebra of linear programs (March 15 and 22, 2015) Many optimization problems can be formulated as linear programs. The main features of a linear program are the following: Variables are real
More informationOutput-sensitive algorithms for sumset and sparse polynomial multiplication
Output-sensitive algorithms for sumset and sparse polynomial multiplication Andrew Arnold Cheriton School of Computer Science University of Waterloo Waterloo, Ontario, Canada Daniel S. Roche Computer Science
More informationCME 302: NUMERICAL LINEAR ALGEBRA FALL 2005/06 LECTURE 6
CME 302: NUMERICAL LINEAR ALGEBRA FALL 2005/06 LECTURE 6 GENE H GOLUB Issues with Floating-point Arithmetic We conclude our discussion of floating-point arithmetic by highlighting two issues that frequently
More informationTotal Ordering on Subgroups and Cosets
Total Ordering on Subgroups and Cosets Alexander Hulpke Department of Mathematics Colorado State University 1874 Campus Delivery Fort Collins, CO 80523-1874 hulpke@math.colostate.edu Steve Linton Centre
More informationarxiv: v2 [cs.sc] 22 Jun 2017
Sparse Polynomial Interpolation with Finitely Many Values for the Coefficients arxiv:1704.04359v [cs.sc] Jun 017 Qiaolong Huang and Xiao-Shan Gao KLMM, UCAS, Academy of Mathematics and Systems Science
More informationEfficient Algorithms for Order Bases Computation
Efficient Algorithms for Order Bases Computation Wei Zhou and George Labahn Cheriton School of Computer Science University of Waterloo, Waterloo, Ontario, Canada Abstract In this paper we present two algorithms
More information4. Parallel computation. a) Processor efficient NC and RNC algorithms for matrix computations with extensions. My paper with
RESEARCH HIGHLIGHTS The results of my research have been published in 3 books and over 230 research articles in journals and refereed proceedings of conferences and have been surveyed in my 18 surveys
More informationRecovery Based on Kolmogorov Complexity in Underdetermined Systems of Linear Equations
Recovery Based on Kolmogorov Complexity in Underdetermined Systems of Linear Equations David Donoho Department of Statistics Stanford University Email: donoho@stanfordedu Hossein Kakavand, James Mammen
More informationLeast Squares Approximation
Chapter 6 Least Squares Approximation As we saw in Chapter 5 we can interpret radial basis function interpolation as a constrained optimization problem. We now take this point of view again, but start
More informationTensor rank-one decomposition of probability tables
Tensor rank-one decomposition of probability tables Petr Savicky Inst. of Comp. Science Academy of Sciences of the Czech Rep. Pod vodárenskou věží 2 82 7 Prague, Czech Rep. http://www.cs.cas.cz/~savicky/
More informationTheorem 5.3. Let E/F, E = F (u), be a simple field extension. Then u is algebraic if and only if E/F is finite. In this case, [E : F ] = deg f u.
5. Fields 5.1. Field extensions. Let F E be a subfield of the field E. We also describe this situation by saying that E is an extension field of F, and we write E/F to express this fact. If E/F is a field
More informationSparse representations from moments
Sparse representations from moments Bernard Mourrain Inria Méditerranée, Sophia Antipolis BernardMourrain@inriafr Sparse representation of signals Given a function or signal f (t): decompose it as f (t)
More informationOutline Introduction: Problem Description Diculties Algebraic Structure: Algebraic Varieties Rank Decient Toeplitz Matrices Constructing Lower Rank St
Structured Lower Rank Approximation by Moody T. Chu (NCSU) joint with Robert E. Funderlic (NCSU) and Robert J. Plemmons (Wake Forest) March 5, 1998 Outline Introduction: Problem Description Diculties Algebraic
More informationCity Research Online. Permanent City Research Online URL:
Christou, D., Karcanias, N. & Mitrouli, M. (2007). A Symbolic-Numeric Software Package for the Computation of the GCD of Several Polynomials. Paper presented at the Conference in Numerical Analysis 2007,
More informationSector-Disk Codes and Partial MDS Codes with up to Three Global Parities
Sector-Disk Codes and Partial MDS Codes with up to Three Global Parities Junyu Chen Department of Information Engineering The Chinese University of Hong Kong Email: cj0@alumniiecuhkeduhk Kenneth W Shum
More informationApproximate GCDs of polynomials and sparse SOS relaxations
Approximate GCDs of polynomials and sparse SOS relaxations Bin Li a, Jiawang Nie b Lihong Zhi a a Key Lab of Mathematics Mechanization, AMSS, Beijing 100190 China b Department of Mathematics, UCSD, La
More informationError estimates for the ESPRIT algorithm
Error estimates for the ESPRIT algorithm Daniel Potts Manfred Tasche Let z j := e f j j = 1,..., M) with f j [ ϕ, 0] + i [ π, π) and small ϕ 0 be distinct nodes. With complex coefficients c j 0, we consider
More informationOn the computation of the Jordan canonical form of regular matrix polynomials
On the computation of the Jordan canonical form of regular matrix polynomials G Kalogeropoulos, P Psarrakos 2 and N Karcanias 3 Dedicated to Professor Peter Lancaster on the occasion of his 75th birthday
More informationSolving Sparse Integer Linear Systems
Solving Sparse Integer Linear Systems Wayne Eberly, Mark Giesbrecht, Giorgi Pascal, Arne Storjohann, Gilles Villard To cite this version: Wayne Eberly, Mark Giesbrecht, Giorgi Pascal, Arne Storjohann,
More informationAdaptive Online Gradient Descent
University of Pennsylvania ScholarlyCommons Statistics Papers Wharton Faculty Research 6-4-2007 Adaptive Online Gradient Descent Peter Bartlett Elad Hazan Alexander Rakhlin University of Pennsylvania Follow
More informationSymbolic-Numeric Sparse Interpolation of Multivariate Polynomials
Symbolic-Numeric Sparse Inerpolaion o Mulivariae Polynomials Mark Giesbrech George Labahn Wen-shin Lee School o Compuer Science Universiy o Waerloo Approimae inerpolaion o polynomials Suppose we can sample
More informationAn Additive Characterization of Fibers of Characters on F p
An Additive Characterization of Fibers of Characters on F p Chris Monico Texas Tech University Lubbock, TX c.monico@ttu.edu Michele Elia Politecnico di Torino Torino, Italy elia@polito.it January 30, 2009
More informationSPARSE signal representations have gained popularity in recent
6958 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 57, NO. 10, OCTOBER 2011 Blind Compressed Sensing Sivan Gleichman and Yonina C. Eldar, Senior Member, IEEE Abstract The fundamental principle underlying
More informationAn Interpolation Algorithm for List Decoding of Reed-Solomon Codes
An Interpolation Algorithm for List Decoding of Reed-Solomon Codes Kwankyu Lee Department of Mathematics San Diego State University San Diego, USA Email: kwankyu@sogangackr Michael E O Sullivan Department
More informationAdvanced Digital Signal Processing -Introduction
Advanced Digital Signal Processing -Introduction LECTURE-2 1 AP9211- ADVANCED DIGITAL SIGNAL PROCESSING UNIT I DISCRETE RANDOM SIGNAL PROCESSING Discrete Random Processes- Ensemble Averages, Stationary
More informationAn Even Order Symmetric B Tensor is Positive Definite
An Even Order Symmetric B Tensor is Positive Definite Liqun Qi, Yisheng Song arxiv:1404.0452v4 [math.sp] 14 May 2014 October 17, 2018 Abstract It is easily checkable if a given tensor is a B tensor, or
More informationOn the use of polynomial matrix approximant in the block Wiedemann algorithm
On the use of polynomial matrix approximant in the block Wiedemann algorithm ECOLE NORMALE SUPERIEURE DE LYON Laboratoire LIP, ENS Lyon, France Pascal Giorgi Symbolic Computation Group, School of Computer
More informationDeterminant evaluations for binary circulant matrices
Spec Matrices 014; :187 199 Research Article Open Access Christos Kravvaritis* Determinant evaluations for binary circulant matrices Abstract: Determinant formulas for special binary circulant matrices
More informationPART II: Research Proposal Algorithms for the Simplification of Algebraic Formulae
Form 101 Part II 6 Monagan, 195283 PART II: Research Proposal Algorithms for the Simplification of Algebraic Formulae 1 Research Area Computer algebra (CA) or symbolic computation, as my field is known
More informationComputational Methods CMSC/AMSC/MAPL 460
Computational Methods CMSC/AMSC/MAPL 460 Fourier transform Balaji Vasan Srinivasan Dept of Computer Science Several slides from Prof Healy s course at UMD Last time: Fourier analysis F(t) = A 0 /2 + A
More informationProbabilistic Algorithms for Computing Resultants
Probabilistic Algorithms for Computing Resultants Michael Monagan Centre for Experimental and Constructive Mathematics Simon Fraser University, Burnaby, B.C. V5A 1S6, CANADA mmonagan@cecm.sfu.ca ABSTRACT
More informationOn the BMS Algorithm
On the BMS Algorithm Shojiro Sakata The University of Electro-Communications Department of Information and Communication Engineering Chofu-shi, Tokyo 182-8585, JAPAN Abstract I will present a sketch of
More informationStatistical Geometry Processing Winter Semester 2011/2012
Statistical Geometry Processing Winter Semester 2011/2012 Linear Algebra, Function Spaces & Inverse Problems Vector and Function Spaces 3 Vectors vectors are arrows in space classically: 2 or 3 dim. Euclidian
More informationMATH 101A: ALGEBRA I, PART D: GALOIS THEORY 11
MATH 101A: ALGEBRA I, PART D: GALOIS THEORY 11 3. Examples I did some examples and explained the theory at the same time. 3.1. roots of unity. Let L = Q(ζ) where ζ = e 2πi/5 is a primitive 5th root of
More informationIndex. book 2009/5/27 page 121. (Page numbers set in bold type indicate the definition of an entry.)
page 121 Index (Page numbers set in bold type indicate the definition of an entry.) A absolute error...26 componentwise...31 in subtraction...27 normwise...31 angle in least squares problem...98,99 approximation
More informationFast Polynomial Multiplication
Fast Polynomial Multiplication Marc Moreno Maza CS 9652, October 4, 2017 Plan Primitive roots of unity The discrete Fourier transform Convolution of polynomials The fast Fourier transform Fast convolution
More informationLattice reduction of polynomial matrices
Lattice reduction of polynomial matrices Arne Storjohann David R. Cheriton School of Computer Science University of Waterloo Presented at the SIAM conference on Applied Algebraic Geometry at the Colorado
More informationCHAPTER 11. A Revision. 1. The Computers and Numbers therein
CHAPTER A Revision. The Computers and Numbers therein Traditional computer science begins with a finite alphabet. By stringing elements of the alphabet one after another, one obtains strings. A set of
More informationThe kernel structure of rectangular Hankel and Toeplitz matrices
The ernel structure of rectangular Hanel and Toeplitz matrices Martin H. Gutnecht ETH Zurich, Seminar for Applied Mathematics ØØÔ»»ÛÛÛºÑ Ø º Ø Þº» Ñ 3rd International Conference on Structured Matrices
More informationAlgorithms for Fast Linear System Solving and Rank Profile Computation
Algorithms for Fast Linear System Solving and Rank Profile Computation by Shiyun Yang A thesis presented to the University of Waterloo in fulfillment of the thesis requirement for the degree of Master
More informationTURNER, WILLIAM JONATHAN. Black Box Linear Algebra with the LinBox Library.
Abstract TURNER, WILLIAM JONATHAN. Black Box Linear Algebra with the LinBox Library. (Under the direction of Erich Kaltofen.) Black box algorithms for exact linear algebra view a matrix as a linear operator
More informationRational Univariate Reduction via Toric Resultants
Rational Univariate Reduction via Toric Resultants Koji Ouchi 1,2 John Keyser 1 Department of Computer Science, 3112 Texas A&M University, College Station, TX 77843-3112, USA Abstract We describe algorithms
More informationScientific Computing
Scientific Computing Direct solution methods Martin van Gijzen Delft University of Technology October 3, 2018 1 Program October 3 Matrix norms LU decomposition Basic algorithm Cost Stability Pivoting Pivoting
More informationNumerical Experiments for Finding Roots of the Polynomials in Chebyshev Basis
Available at http://pvamu.edu/aam Appl. Appl. Math. ISSN: 1932-9466 Vol. 12, Issue 2 (December 2017), pp. 988 1001 Applications and Applied Mathematics: An International Journal (AAM) Numerical Experiments
More informationTowards High Performance Multivariate Factorization. Michael Monagan. This is joint work with Baris Tuncer.
Towards High Performance Multivariate Factorization Michael Monagan Center for Experimental and Constructive Mathematics Simon Fraser University British Columbia This is joint work with Baris Tuncer. The
More informationGeometric Modeling Summer Semester 2010 Mathematical Tools (1)
Geometric Modeling Summer Semester 2010 Mathematical Tools (1) Recap: Linear Algebra Today... Topics: Mathematical Background Linear algebra Analysis & differential geometry Numerical techniques Geometric
More information