Conference PaperPDF Available

Abstract

The computation of matrix trigonometric functions has received remarkable attention in the last decades due to its usefulness in the solution of systems of second order linear differential equations. Recently, several state-of-the-art algorithms have been provided for computing these matrix functions, in particular for the matrix cosine function.
Modelling for Engineering
& Human Behaviour 2019
Val`encia, 10 12 July 2019
This book includes the extended abstracts of papers presented at XXIst Edition of the Mathe-
matical Modelling Conference Series at the Institute for Multidisciplinary Mathematics “Math-
ematical Modelling in Engineering & Human Behaviour”.
I.S.B.N.: 978-84-09-16428-8
Version: 18/11/19
Report any problems with this document to ellona1@upvnet.upv.es.
Edited by: R. Company, J. C. Cort´es, L. J´odar and E. L´opez-Navarro.
Credits: The cover has been designed using images from kjpargeter/freepik.
Instituto Universitario de Matem´atica
Multidisciplinar
This book has been supported by the European
Union through the Operational Program of the [Eu-
ropean Regional Development Fund (ERDF) / Eu-
ropean Social Fund (ESF)] of the Valencian Com-
munity 2014-2020. [Record: GJIDI/2018/A/010].
Contents
A personality mathematical model of placebo with or without deception: an application
of the Self-Regulation Therapy..........................................................1
The role of police deterrence in urban burglary prevention: a new mathematical ap-
proach..................................................................................9
A Heuristic optimization approach to solve berth allocation problem . . . . . . . . . . . . . . . . .14
Improving the efficiency of orbit determination processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
A new three-steps iterative method for solving nonlinear systems . . . . . . . . . . . . . . . . . . . . 22
Adaptive modal methods to integrate the neutron diffusion equation . . . . . . . . . . .. . . . . . 26
Numerical integral transform methods for random hyperbolic models . . . . . . . . . . . . . . . . 32
Nonstandard finite difference schemes for coupled delay differential models . . . . . . . . . . .37
Semilocal convergence for new Chebyshev-type iterative methods . . . . . . . . . . . . . . . . . . . 42
Mathematical modeling of Myocardial Infarction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Symmetry relations between dynamical planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Econometric methodology applied to financial systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .56
New matrix series expansions for the matrix cosine approximation . . . . . . . . . . . . . . . . . . . 64
Modeling the political corruption in Spain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Exponential time differencing schemes for pricing American option under the Heston
model................................................................................. 75
Chromium layer thickness forecast in hard chromium plating process using gradient
boosted regression trees: a case study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Design and convergence of new iterative methods with memory for solving nonlinear
problems ..............................................................................83
Study of the influence falling friction on the wheel/rail contact in railway dynamics . . 88
Extension of the modal superposition method for general damping applied in railway
dynamics..............................................................................94
Predicting healthcare cost of diabetes using machine learning models . . . . . . . . . . . . . . . . 99
iv
Modelling for Engineering & Human Behaviour 2019
Sampling of pairwise comparisons in decision-making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
A multi-objective and multi-criteria approach for district metered area design: water
operation and quality analysis ........................................................110
Updating the OSPF routing protocol for communication networks by optimal decision-
making over the k-shortest path algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Optimal placement of quality sensors in water distribution systems . . . . . . . . . . . . . . . . . 124
Mapping musical notes to socio-political events. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Comparison between DKGA optimization algorithm and Grammar Swarm surrogated
model applied to CEC2005 optimization benchmark. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
The quantum brain model ......................................................... 142
Probabilistic solution of a randomized first order differential equation with discrete
delay.................................................................................151
A predictive method for bridge health monitoring under operational conditions . . . . . 155
Comparison of a new maximum power point tracking based on neural network with
conventional methodologies ...........................................................160
Influence of different pathologies on the dynamic behaviour and against fatigue of
railway steel bridges..................................................................166
Statistical-vibratory analysis of wind turbine multipliers under different working con-
ditions................................................................................171
Analysis of finite dimensional linear control systems subject to uncertainties via prob-
abilistic densities ..................................................................... 176
Topographic representation of cancer data using Boolean Networks . . . . . . . . . . . . . . . . . 180
Trying to stabilize the population and mean temperature of the World . . . . . . . . . . . . . 185
Optimizing the demographic rates to control the dependency ratio in Spain . . . . . . . . .193
An integer linear programming approach to check the embodied CO2emissions of the
opaque part of a fa¸cade...............................................................199
Acoustics on the Poincar´e Disk.....................................................206
Network computational model to estimate the effectiveness of the influenza vaccine a
posteriori .............................................................................211
The key role of Liouville-Gibbs equation for solving random differential equations:
Some insights and applications........................................................217
v
13
New matrix series expansions for the matrix cosine
approximation
Emilio Defez [1, Javier Ib´nez\, Jos´e M. Alonso\, Jes´us Peinado]and Pedro Alonso-Jord´a
([) Institut Universitari de Matem`atica Multidisciplinar,
Universitat Polit`ecnica de Val`encia,
(\) Instituto de Instrumentaci´on para Imagen Molecular,
Universitat Polit`ecnica de Val`encia,
(]) Departamento de Sistemas Inform´aticos y Computaci´on,
Universitat Polit`ecnica de Val`encia,
() Grupo Interdisciplinar de Computaci´on y Comunicaciones,
Universitat Polit`ecnica de Val`encia.
1 Introduction and notation
The computation of matrix trigonometric functions has received remarkable attention in the
last decades due to its usefulness in the solution of systems of second order linear differential
equations. Recently, several state-of-the-art algorithms have been provided for computing these
matrix functions, see [1–4], in particular for the matrix cosine function.
Among the proposed methods for the approximate computation of the matrix cosine, two fun-
damental ones stand out: those based on rational approximations [1,5–7], and those related to
polynomial approximations, using either Taylor series developments [8,9] or serial developments
of Hermite matrix polynomials [10]. In general, polynomial approximations showed to be more
efficient than the rational algorithms in tests because they are more accurate despite a slightly
higher cost.
Bernoulli polynomials and Bernoulli numbers have been extensively used in several areas of
mathematics (an excelent survey about Bernoulli polynomials and its applicacions can be found
in [11]).
In this paper, we will present a new series development of the matrix cosine in terms of the
Bernoulli matrix polynomials. We are going to verify that its use allows obtaining a new and
competitive method for the approximation of the matrix cosine.
The organization of the paper is as follows: In Section 2, we will obtain two serial developments
of the matrix cosine in terms of the Bernoulli matrix polynomials. In Section 3, we will present
1e-mail: edefez@imm.upv.es
64
Modelling for Engineering & Human Behaviour 2019
the different numerical tests performed. Conclusions are given in Section 4.
Throughout this paper, we denote by Cr×rthe set of all the complex square matrices of size
r. Besides, we denote Ias the identity matrix in Cr×r. A polynomial of degree mis given by
an expression of the form Pm(t) = amtm+am1tm1+··· +a1t+a0, where tis a real variable
and aj, for 0 jm, are complex numbers. Moreover, we can define the matrix polynomial
Pm(B) for BCr×ras Pm(B) = amBm+am1Bm1+· ·· +a1B+a0I. As usual, the matrix
norm ··k denotes any subordinate matrix norm; in particular ··k1is the usual 1norm.
2 On Bernoulli matrix polynomials
The Bernoulli polynomials Bn(x) are defined in [12, p.588] as the coefficients of the generating
function
g(x, t) = tetx
et1=X
n0
Bn(x)
n!tn,|t|<2π, (1)
where g(x, t) is an holomorphic function in Cfor the variable t(it has an avoidable singularity
in t= 0). Bernoulli polynomials Bn(x) has the explicit expression
Bn(x) =
n
X
k=0 n
k!Bkxnk,(2)
where the Bernoulli numbers are defined by Bn=Bn(0). Therefore, it follows that the Bernoulli
numbers satisfy
z
ez1=X
n0
Bn
n!zn,|z|<2π, (3)
where
B0= 1, Bk=
k1
X
i=0 k
i!Bi
k+ 1 i, k 1.(4)
Note that B3=B5=··· =B2k+1 = 0, for k1. For a matrix ACr×r, we define the mth
Bernoulli matrix polynomial by the expression
Bm(A) =
m
X
k=0 m
k!BkAmk.(5)
We can use the series expansion
eAt = et1
t!X
n0
Bn(A)tn
n!,|t|<2π, (6)
to obtain approximations of the matrix exponential. A method based in (6) to approximate
the exponential matrix has been presented in [13].
65
Modelling for Engineering & Human Behaviour 2019
From (6), we obtain the following expression for the matrix cosine and sine:
cos (A) = (cos (1) 1) X
n0
(1)nB2n+1(A)
(2n+ 1)! + sin (1) X
n0
(1)nB2n(A)
(2n)! ,
sin (A) = sin (1) X
n0
(1)nB2n+1(A)
(2n+ 1)! (cos (1) 1) X
n0
(1)nB2n(A)
(2n)! .
(7)
Note that unlike the Taylor (and Hermite) polynomials that are even or odd, depending on the
parity of the polynomial degree n, the Bernoulli polynomials do not verify this property. Thus,
in the development of cos(A) and sin (A), all Bernoulli polynomials are needed (and not just
the even-numbered ones).
Replacing in (6) the value tfor it and it respectively and taking the arithmetic mean, we
obtain the expression
X
n0
(1)nB2n(A)
(2n)! t2n=t
2 sin t
2cos tA t
2I ,|t|<2π. (8)
Taking t= 2 in (8) it follows that
cos (A) = sin (1) X
n0
(1)n22nB2nA+I
2
(2n)! ,(9)
Note that in formula (9) only even grade Bernoulli’s polynomials appear.
3 Numerical Experiments
Having in mind expressions (7) and (9), two different approximations are given to compute
cosine matrix function.
To test the proposed method and the two distinct approximations, and to compare them with
other approaches, the following algorithms have been implemented on MATLAB R2018b:
-cosmber. New code based on the new developments of Bernoulli matrix polynomials (formulae
(7) and (9)). The maximum value of mto be used is m= 36, with even and odd terms.
-cosmtay. Code based on the Taylor series for the cosine [8]. It will provide a maximum value
of m= 16, considering only the even terms, which would be equivalent to m= 32 using
the even and odd terms.
-cosmtayher. Code based on the Hermite series for the cosine [10]. As mentioned before, it
will provide a maximum value of m= 16.
-cosm. Code based on the Pad´e rational approximation for the cosine [7].
66
Modelling for Engineering & Human Behaviour 2019
The following sets of matrices have been used:
a) Diagonalizable matrices. The matrices have been obtained as A=V·D·VT, where D
is a diagonal matrix (with complex or real values) and matrix Vis an orthogonal matrix,
V=H/16, where His a Hadamard matrix. We have 2.18 ≤ kAk1207.52. The matrix
cosine is exactly calculated as cos (A) = V·cos (D)·VT.
b) Non-diagonalizables matrices. The matrices have been computed as A=V·J·V1,
where Jis a Jordan matrix with complex eigenvalues with module less than 10 and random
algebraic multiplicity between 1 and 5. Matrix Vis a random matrix with elements in
the interval [0.5,0.5]. We have 1279.16 ≤ kAk187886.4. The matrix cosine is exactly
calculated as cos (A) = V·cos (J)·V1.
c) Matrices from the Matrix Computation Toolbox [14] and from the Eigtool Matlab
package [15]. These matrices have been chosen because they have more varied and
significant characteristics.
In the numerical test, we used 259 matrices of size 128 ×128: 100 from the diagonalizable
set, 100 from the non-diagonalizable set, 42 from Matrix Computation Toolbox and 17 from
Eigtool Matlab package. Results are given in Tables 1 and 2. The rows of each table show the
percentage of cases in which the relative errors of cosmber (Bernoulli) is lower, greater or equal
than the relative errors of cosmtay (Taylor), cosmtayher (Hermite) and cosm (Pad´e). Graphics
of the Normwise relative errors and the Performance Profile are given in Figures 1 and 2. The
total number of matrix products was: 3202 (cosmber ), 2391 (cosmtay), 1782 (cosmtayher) and
3016 (cosm). Recall that in the Bernoulli implementation, the maximum value of mto be used
was m= 36 considering all the terms and, in the rest of algorithms, was m= 32 but just
having into account the even terms.
E(cosmber)< E(cosmtay) 55.60%
E(cosmber)> E(cosmtay) 44.40%
E(cosmber) = E(cosmtay) 0%
E(cosmber)< E(cosmtayher) 50.97%
E(cosmber)> E(cosmtayher) 49.03%
E(cosmber) = E(cosmtayher) 0%
E(cosmber)< E(cosm) 76.83%
E(cosmber)> E(cosm) 23.17%
E(cosmber) = E(cosm) 0%
Table 1: Using approximation (7)
E(cosmber)< E(cosmtay) 65.64%
E(cosmber)> E(cosmtay) 34.36%
E(cosmber) = E(cosmtay) 0%
E(cosmber)< E(cosmtayher) 60.62%
E(cosmber)> E(cosmtayher) 39.38%
E(cosmber) = E(cosmtayher) 0%
E(cosmber)< E(cosm) 73.75%
E(cosmber)> E(cosm) 26.25%
E(cosmber) = E(cosm) 0%
Table 2: Using approximation (9)
67
Modelling for Engineering & Human Behaviour 2019
0 50 100 150 200 250 300
Matrix
10-150
10-100
10-50
100
Er
cond*u
cosmber
cosmtay
cosmtayher
cosm
(a) Using formula (7)
0 50 100 150 200 250 300
Matrix
10-150
10-100
10-50
100
Er
cond*u
cosmber
cosmtay
cosmtayher
cosm
(b) Using formula (9)
Figure 1: Normwise relative errors.
12345
0
0.2
0.4
0.6
0.8
1
p
cosmber
cosmtay
cosmtayher
cosm
(a) Using formula (7)
12345
0
0.2
0.4
0.6
0.8
1
p
cosmber
cosmtay
cosmtayher
cosm
(b) Using formula (9)
Figure 2: Performance Profile.
4 Conclusions
In general, the implementation based on the new Bernoulli series (9) is more accurate than (7),
comparing it with the one based on the Taylor series, algorithm (cosmtay) and Hermite series,
algorithm (cosmtayher), and the one based in Pad´e rational approximation, algorithm (cosm).
Acknowledgement
This work has been partially supported by Spanish Ministerio de Econom´ıa y Competitividad
and European Regional Development Fund (ERDF) grants TIN2017-89314-P and by the Pro-
grama de Apoyo a la Investigaci´on y Desarrollo 2018 of the Universitat Polit`ecnica de Val`encia
(PAID-06-18) grants SP20180016.
References
[1] Serbin, S.M. and Blalock, S.A., An algorithm for computing the matrix cosine. SIAM
Journal on Scientific Computing, 1(2): 198–204, 1980.
68
Modelling for Engineering & Human Behaviour 2019
[2] Dehghan, M. and Hajarian, M., Computing matrix functions using mixed interpolation
methods. Mathematical and Computer Mod- el ling, 52(5-6): 826–836, 2010.
[3] Higham, N.J., Functions of Matrices: Theory and Computation. SIAM, Philadelphia, PA,
USA, 2008.
[4] Alonso-Jord´a, P., Peinado, J., Ib´a˜nez, J., Sastre, J. and Defez, E., Computing matrix
trigonometric functions with GPUs through Matlab. The Journal of Supercomputing, pages
1–14, 2018.
[5] Tsitouras, Ch. and Katsikis, V., Bounds for variable degree ratio- nal Lapproximations
to the matrix cosine. Computer Physics Communications, 185(11): 2834–2840, 2014.
[6] Serbin, S.M., Rational approximations of trigonometric matrices with application to
second-order systems of differential equations. Applied Mathematics and Computation,
5(1): 75–92, 1979.
[7] Al-Mohy, A.H., Higham, N.J. and Relton, S.D., New algorithms for computing the matrix
sine and cosine separately or simultaneously. SIAM Journal on Scientific Computing, 37(1):
A456–A487, 2015.
[8] Sastre, J., Ib´nez, J., Alonso-Jord´a, P., Peinado, J. and Defez, E., Two algorithms for com-
puting the matrix cosine function. Applied Mathe- matics and Computation, 312: 66–77,
2017.
[9] Sastre, J., Ib´nez, J., Alonso-Jord´a, P., Peinado, J. and Defez, E., Fast Taylor polynomial
evaluation for the computation of the matrix cosine. Journal of Computational and Applied
Mathematics, 354: 641–650, 2019.
[10] Defez, E., Ib´nez, J., Peinado, J., Sastre, J. and Alonso-Jord´a, P., An efficient and accurate
algorithm for computing the matrix cosine based on new Hermite approximations. Journal
of Computational and Applied Mathematics, 348: 1–13, 2019.
[11] Kouba, O., Lecture Notes, Bernoulli Polynomials and Applications. arXiv preprint
arXiv:1309.7560, 2013.
[12] WJ Olver, F., W Lozier, D., F Boisvert, R. and W Clark, C., NIST handbook of mathe-
matical functions hardback and CD-ROM. Cambridge University Press, 2010.
[13] Defez, E., Ib´nez, J., Peinado, J., Alonso-Jord´a, P. and Alonso, J.M., Computing matrix
functions by matrix Bernoulli series. In 19th International Conference on Computational
and Mathematical Methods in Science and Engineering (CMMSE-2019), From 30th of Juny
to 6th of July 2019. Poster presented at some conference, Rota (C´adiz), Spain.
[14] Higham, N.J., The test matrix toolbox for MATLAB (Version 3.0). University of Manch-
ester Manchester, 1995.
[15] Wright, TG., Eigtool, version 2.1. URL: web.comlab.ox.ac.uk/pseudospectra/eigtool,
2009.
69
ResearchGate has not been able to resolve any citations for this publication.
Conference Paper
Full-text available
Bernoulli polynomials and Bernoulli numbers have been extensively used in several areas of mathematics (especially in number theory) because they appear in many mathematical formulas: as in the residual term of the Euler-Maclaurian quadrature rule, in the Taylor series expansion of the trigonometric functions tan (x), csc (x) and cot (x), in the Taylor series expansion of the hyperbolic function tanh(x). They also appear in the well known exact expression of the even integer values of the Riemann zeta function.
Article
Full-text available
In this work we introduce a new method to compute the matrix cosine. It is based on recent new matrix polynomial evaluation methods for the Taylor approximation and a mixed forward and backward error analysis. The matrix polynomial evaluation methods allow to evaluate the Taylor polynomial approximation of the matrix cosine function more efficiently than using Paterson-Stockmeyer method. A sequential Matlab implementation of the new algorithm is provided, giving better efficiency and accuracy than state-of-the-art algorithms. Moreover, we provide an implementation in Matlab that can use NVIDIA GPUs easily and efficiently.
Article
Full-text available
In this work we introduce new rational-polynomial Hermite matrix expansions which allow us to obtain a new accurate and efficient method for computing the matrix cosine. This method is compared with other state-of-the-art methods for computing the matrix cosine, including a method based on Padé approximants, showing a far superior efficiency, and higher accuracy. The algorithm implemented on the basis of this method can also be executed either in one or two NVIDIA GPUs, which demonstrates its great computational capacity.
Article
Full-text available
This paper presents an implementation of one of the most up-to-day algorithms proposed to compute the matrix trigonometric functions sine and cosine. The method used is based on Taylor series approximations which intensively uses matrix multiplications. To accelerate matrix products, our application can use from one to four NVIDIA GPUs by using the NVIDIA cublas and cublasXt libraries. The application, implemented in C++, can be used from the Matlab command line thanks to the mex files provided. We experimentally assess our implementation in modern and very high-performance NVIDIA GPUs.
Article
Full-text available
The computation of matrix trigonometric functions has received remarkable attention in the last decades due to its usefulness in the solution of systems of second order linear differential equations. Several state-of-the-art algorithms have been provided recently for computing these matrix functions. In this work, we present two efficient algorithms based on Taylor series with forward and backward error analysis for computing the matrix cosine. A MATLAB implementation of the algorithms is compared to state-of-the-art algorithms, with excellent performance in both accuracy and cost.
Article
Full-text available
Several existing algorithms for computing the matrix cosine employ polynomial or rational approximations combined with scaling and use of a double angle formula. Their derivations are based on forward error bounds. We derive new algorithms for computing the matrix cosine, the matrix sine, and both simultaneously that are backward stable in exact arithmetic and behave in a forward stable manner in floating point arithmetic. Our new algorithms employ both Pade approximants of sin x and new rational approximants to cos x and sin x obtained from Pade approximants to e(x). The amount of scaling and the degree of the approximants are chosen to minimize the computational cost subject to backward stability in exact arithmetic. Numerical experiments show that the new algorithms have backward and forward errors that rival or surpass those of existing algorithms and are particularly favorable for triangular matrices.
Article
Full-text available
Modern developments in theoretical and applied science depend on knowledge of the properties of mathematical functions, from elementary trigonometric functions to the multitude of special functions. These functions appear whenever natural phenomena are studied, engineering problems are formulated, and numerical simulations are performed. They also crop up in statistics, financial models, and economic analysis. Using them effectively requires practitioners to have ready access to a reliable collection of their properties. This handbook results from a 10-year project conducted by the National Institute of Standards and Technology with an international group of expert authors and validators. Printed in full color, it is destined to replace its predecessor, the classic but long-outdated Handbook of Mathematical Functions, edited by Abramowitz and Stegun. Included with every copy of the book is a CD with a searchable PDF of each chapter.
Article
A thorough and elegant treatment of the theory of matrix functions and numerical methods for computing them, including an overview of applications, new and unpublished research results, and improved algorithms. Key features include a detailed treatment of the matrix sign function and matrix roots; a development of the theory of conditioning and properties of the Frechet derivative; Schur decomposition; block Parlett recurrence; a thorough analysis of the accuracy, stability, and computational cost of numerical methods; general results on convergence and stability of matrix iterations; and a chapter devoted to the f(A)b problem. Ideal for advanced courses and for self-study, its broad content, references and appendix also make this book a convenient general reference. Contains an extensive collection of problems with solutions and MATLAB implementations of key algorithms.