DKA method for single variable holomorphic functions

Similar documents
Lecture 12 APPROXIMATION OF FIRST ORDER DERIVATIVES

CHAPTER VI Statistical Analysis of Experimental Data

Introduction to local (nonparametric) density estimation. methods

Beam Warming Second-Order Upwind Method

( ) 2 2. Multi-Layer Refraction Problem Rafael Espericueta, Bakersfield College, November, 2006

1 0, x? x x. 1 Root finding. 1.1 Introduction. Solve[x^2-1 0,x] {{x -1},{x 1}} Plot[x^2-1,{x,-2,2}] 3

UNIT 2 SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS

Mu Sequences/Series Solutions National Convention 2014

PGE 310: Formulation and Solution in Geosystems Engineering. Dr. Balhoff. Interpolation

Derivation of 3-Point Block Method Formula for Solving First Order Stiff Ordinary Differential Equations

MULTIDIMENSIONAL HETEROGENEOUS VARIABLE PREDICTION BASED ON EXPERTS STATEMENTS. Gennadiy Lbov, Maxim Gerasimov

A Remark on the Uniform Convergence of Some Sequences of Functions

1 Lyapunov Stability Theory

Strong Convergence of Weighted Averaged Approximants of Asymptotically Nonexpansive Mappings in Banach Spaces without Uniform Convexity

Unsupervised Learning and Other Neural Networks

C-1: Aerodynamics of Airfoils 1 C-2: Aerodynamics of Airfoils 2 C-3: Panel Methods C-4: Thin Airfoil Theory

Block-Based Compact Thermal Modeling of Semiconductor Integrated Circuits

02/15/04 INTERESTING FINITE AND INFINITE PRODUCTS FROM SIMPLE ALGEBRAIC IDENTITIES

Chapter 5 Properties of a Random Sample

The Mathematical Appendix

Point Estimation: definition of estimators

L5 Polynomial / Spline Curves

BERNSTEIN COLLOCATION METHOD FOR SOLVING NONLINEAR DIFFERENTIAL EQUATIONS. Aysegul Akyuz Dascioglu and Nese Isler

1 Onto functions and bijections Applications to Counting

Functions of Random Variables

On Modified Interval Symmetric Single-Step Procedure ISS2-5D for the Simultaneous Inclusion of Polynomial Zeros

Fractional Order Finite Difference Scheme For Soil Moisture Diffusion Equation And Its Applications

Estimation of Stress- Strength Reliability model using finite mixture of exponential distributions

1. A real number x is represented approximately by , and we are told that the relative error is 0.1 %. What is x? Note: There are two answers.

On the convergence of derivatives of Bernstein approximation

Arithmetic Mean and Geometric Mean

Summary of the lecture in Biostatistics

Random Variate Generation ENM 307 SIMULATION. Anadolu Üniversitesi, Endüstri Mühendisliği Bölümü. Yrd. Doç. Dr. Gürkan ÖZTÜRK.

Evolution Operators and Boundary Conditions for Propagation and Reflection Methods

Chapter 4 (Part 1): Non-Parametric Classification (Sections ) Pattern Classification 4.3) Announcements

X ε ) = 0, or equivalently, lim

PTAS for Bin-Packing

2006 Jamie Trahan, Autar Kaw, Kevin Martin University of South Florida United States of America

Analysis of Lagrange Interpolation Formula

Solving Constrained Flow-Shop Scheduling. Problems with Three Machines

Numerical Simulations of the Complex Modied Korteweg-de Vries Equation. Thiab R. Taha. The University of Georgia. Abstract

18.413: Error Correcting Codes Lab March 2, Lecture 8

Generalization of the Dissimilarity Measure of Fuzzy Sets

Lecture 5: Interpolation. Polynomial interpolation Rational approximation

Algebraic Condition for Integrable Numerical Algorithms

Cubic Nonpolynomial Spline Approach to the Solution of a Second Order Two-Point Boundary Value Problem

x y exp λ'. x exp λ 2. x exp 1.

PROJECTION PROBLEM FOR REGULAR POLYGONS

Lecture 07: Poles and Zeros

ABOUT ONE APPROACH TO APPROXIMATION OF CONTINUOUS FUNCTION BY THREE-LAYERED NEURAL NETWORK

7.0 Equality Contraints: Lagrange Multipliers

Investigating Cellular Automata

Comparing Different Estimators of three Parameters for Transmuted Weibull Distribution

MMJ 1113 FINITE ELEMENT METHOD Introduction to PART I

Transforms that are commonly used are separable

best estimate (mean) for X uncertainty or error in the measurement (systematic, random or statistical) best

Research Article A New Iterative Method for Common Fixed Points of a Finite Family of Nonexpansive Mappings

Lecture 9: Tolerant Testing

On the Interval Zoro Symmetric Single Step. Procedure IZSS1-5D for the Simultaneous. Bounding of Real Polynomial Zeros

Numerical Analysis Formulae Booklet

1 Convergence of the Arnoldi method for eigenvalue problems

A tighter lower bound on the circuit size of the hardest Boolean functions

CHAPTER 4 RADICAL EXPRESSIONS

EVALUATION OF FUNCTIONAL INTEGRALS BY MEANS OF A SERIES AND THE METHOD OF BOREL TRANSFORM

Fourth Order Four-Stage Diagonally Implicit Runge-Kutta Method for Linear Ordinary Differential Equations ABSTRACT INTRODUCTION

Analysis of Variance with Weibull Data

CS5620 Intro to Computer Graphics

Extreme Value Theory: An Introduction

A Collocation Method for Solving Abel s Integral Equations of First and Second Kinds

Sampling Theory MODULE V LECTURE - 14 RATIO AND PRODUCT METHODS OF ESTIMATION

means the first term, a2 means the term, etc. Infinite Sequences: follow the same pattern forever.

LINEARLY CONSTRAINED MINIMIZATION BY USING NEWTON S METHOD

Discrete Mathematics and Probability Theory Fall 2016 Seshia and Walrand DIS 10b

MATH 247/Winter Notes on the adjoint and on normal operators.

A conic cutting surface method for linear-quadraticsemidefinite

Ideal multigrades with trigonometric coefficients

STK4011 and STK9011 Autumn 2016

( ) = ( ) ( ) Chapter 13 Asymptotic Theory and Stochastic Regressors. Stochastic regressors model

Q-analogue of a Linear Transformation Preserving Log-concavity

KLT Tracker. Alignment. 1. Detect Harris corners in the first frame. 2. For each Harris corner compute motion between consecutive frames

An Introduction to. Support Vector Machine

h-analogue of Fibonacci Numbers

A new type of optimization method based on conjugate directions

LINEAR REGRESSION ANALYSIS

arxiv: v4 [math.nt] 14 Aug 2015

Binary classification: Support Vector Machines

Part 4b Asymptotic Results for MRR2 using PRESS. Recall that the PRESS statistic is a special type of cross validation procedure (see Allen (1971))

Multivariate Transformation of Variables and Maximum Likelihood Estimation

Chapter 9 Jordan Block Matrices

d dt d d dt dt Also recall that by Taylor series, / 2 (enables use of sin instead of cos-see p.27 of A&F) dsin

Extend the Borel-Cantelli Lemma to Sequences of. Non-Independent Random Variables

Integral Equation Methods. Jacob White. Thanks to Deepak Ramaswamy, Michal Rewienski, Xin Wang and Karen Veroy

CSE 5526: Introduction to Neural Networks Linear Regression

CS 1675 Introduction to Machine Learning Lecture 12 Support vector machines

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model

Lecture Note to Rice Chapter 8

0/1 INTEGER PROGRAMMING AND SEMIDEFINTE PROGRAMMING

Chapter 5. Curve fitting

Fibonacci Identities as Binomial Sums

Bootstrap Method for Testing of Equality of Several Coefficients of Variation

Transcription:

DKA method for sgle varable holomorphc fuctos TOSHIAKI ITOH Itegrated Arts ad Natural Sceces The Uversty of Toushma -, Mamhosama, Toushma, 770-8502 JAPAN Abstract: - Durad-Kerer-Aberth (DKA method for sgle varable holomorphc fuctos are treated. Fte seres appromato of the sgle varable holomorphc fucto s used to fd local roots of sgle varable holomorphc fuctos stead of algebrac equato. We ca fd roots of sgle varable holomorphc fuctos wde rego by cotug local solutos. Key-Words: - Durad-Kerer-Aberth (DKA method, root fdg Itroducto Advatage of Newto s terato method for o-lear equato or algebrac equato s well-ow ad detaled eplaatos ad ts applcatos are foud may publcatos []. However, brdge for the gap betwee algebrac equato ad aalytc fucto wth ths method was ot gvg much. To cover ths gap [2] s wored towards costructg mathematcal base. Here we wll attempt to cover ths gap actual problem. I ths artcle, root fdg method DKA [3, 4] ad 3rd order methods are treated from both algebrac ad aalytc pots. From ths pot, DKA method resembles to Euclda elmato (algebra ad Weerstrass preparato theorem (aalyss. Numercal algorthms are ot well vestgated from above med pot up to ths pot. We epect to obta ew sght from ths pot.. DKA method ad 3rd order method Durad -Kerer -Aberth method s umercal root fdg algorthm for algebrac equato. Cosder -th degree algebrac equato wth real umber coeffcet, ( P( + a + + a 0, a 0, C. The umber of roots ca be foud umercally by followg Newto s method, (2 ( ( f(,, f(, f (, J ( ( f ( /, ( f(,, f( J( f (, (3 : terato umber ad ϕ (,,...,, m, ϕ, m 2 2 m 0 < < < m f(, 2,..., ( ϕ (, 2,..., a 0,,2,...,. The (3 ca be wrtte as (4 ( ( ( P( / (,,2,...,. We ca assume that (4 s holomorphc mappg, f ( 0 at every terato step wth proper tal codtos. It s a acceptable assumpto, because (0 tal codtos for are gve o the crcle usually. Therefore (4 maps tal crcle where tal values are gve to the closed curve o that eact roots. It s ow that such uque holomorphc fucto that maps crcle to closed curve C ests. From algebrac geometrcal pot, (4 gves umbers of geerators for deals, because each ( + equato (4 s depedet. We ca troduce ope coverg whch gves restrcto mappg, ad etc., D D, D D φ, ( + ρ : D D (, ϕ : +, see Fg.. ( + (0 ( ( D ( D Fg. DKA method ad rego of the mappg. ( D + ( Z ( + D ( D ( D

Ths property may gve superor coverget property. DKA method (4 ca be deformed as, (5 ( + P ( ( (,,2,...,. It resembles to followg Euclda elmato, (6 P ( ( h( + r(,,2,...,, here h( (. I addto r( 0,,...,, the (6 becomes DKA method. DKA method seems umercal appromato of Euclda elmato. Sce t s ow that polyomal fuctos C are UFD, estece of umercal soluto of the method s guarateed. The followg method s modfcato of DKA method, ad t s called 3rd order method, (7 ( ( P ( /{ P'( P ( },,2,...,. 2 DKA method for sgle varable holomorphc fucto We ca modfy DKA method for holomorphc ( fucto f( w, wth several varables, (8 u ( w, f ( w, / (,,2,...,. 2 + P( w, u ( w, f ( w,, w {,,...,,,..., } P w (, ( s Weerstrass polyomal by, ad u ( w, are uts. Estece of such dvso Puf s guarateed by Weerstrass preparato theorem [5]. Fgure 2 llustrates mage of ut U(. Holomorphc fuctos coverge the covergece doma of radus r, (uformly ad absolutely coverge wth approprate radus s<r, see Apped also, (9 f( a, a C, 0 U( here Cauchy-Hadamard theorem for r, 3 2.5 2.5 0.5 Z - 2-0.5 Fg. 2 uw (,, Pw (, fw (, ( lm sup a, r ( ( a. r We troduce varable trasform, ad appromate f( by fte seres, rw (, rw (, :t <, (2 f ( f ( frw ( ( ar ( w bw, b ar (. 0 0 The r of fw ( respect to coeffcets b ad depedet varable w roughly equals to, ad r( s polyradus that depeds o f ad. For eample, appromato by fte seres epaso of Bessel fucto, ths case r(209.05746, behaves as Fg.3. - - 20 40 60 80 00 BesselJ 0( BesselJ 0( r(20 2 4 6 8 0 20 order seres appromato Fg. 3 Comparso betwee orgal fucto ad appromato fucto by fte seres epaso We wll use polyomal fucto by fte seres P ( f ( stead of f( wth DKA ad 3rd order method to resolve local roots of f(. 2. Uform property of holomorphc (aalytc fucto Frstly, we should cofrm uform property of the appromato fuctos by fte seres (polyomals. Fortuately, uform property of the appromato fucto s preserved the approprate treatmet of the radus of the polydsc of orgal holomorphc fucto. If we use r( whch s suffcetly smaller tha radus of covergece, appromato fucto closes orgal fucto, because both fuctos are holomorphc fuctos whch uformly coverge the polydsc of r(. Fgure 4 shows behavor of r( respect to for ep(. The fgure shows that covergece radus r( of the appromato fucto becomes large by

creasg. Sce ar ( O( as to appromato fucto, therefore polyomal fucto for appromato dverges where > r (. I geeral we eed careful treatmet for r( respect to the meag of lm sup (. 2.2 Covergece radus ad ts behavor by fte seres appromato Wth appromate radus r(, we troduce varable trasform r(w. Dscardg the term of + ad hghers the seres epaso of f(, the appromato fucto s, (3 r( f ( w f ( w ar ( w, 0 the, Trucato error Oa ( r + + (. Sce the covergece radus r by ew coeffcets b a r( roughly equals to oe, we oly tae up resolved roots by the DKA whch are sde ut crcle wth followg error correcto. We defe, ( ε f ( ( ( ( 0, δ f(, ad γ f( f (, Corrected radus (trustable, (4 r r ( r, correct error ma ε, δ, γ, γ rerror,,..., m ( ( ( ( f '( f '( f '( f '( ( ( here rw (, lower de s the umber of roots ad m. r 35 30 25 20 5 0 5 f f(e 20 40 60 80 00 Fg. 4 Plot of r( ε f ( '( ε ( Fg. 5 Error estmato ad correcto of radus γ δ f ( f( Easly we ca do error estmato by whether δ < Tolerace Error s satsfed or ot for the resolved roots. If there are ay resolved roots that do ot satsfy ths codto, they are reected. Moreover, resolved roots outsde of the polydsc radus of whch s defed as dstace betwee the ceter of covergece ad reected resolved root wth (4 are also reected. Ths frst process decreases comple error estmato wth comple error treatmet. 2.3 Normalg requremet for actual eecuto by fte seres appromato We defed polyomal fucto for appromato as (2. Applg DKA method to f ( w (2, we must ormale t as P( (. The we devde f ( w by b for ormalato. It s mportat to avod dvergece of coeffcets of polyomal fucto for appromato. We foud that varable trasform (2 removes ths dvergece. If we scamp ths treatmet, we eed large dgt for computato. Because a s very small for large, we eed large dgt for / a for ormalato. 3 Eamples 3. Low order polyomal As a eample, we appled the method to BesselJ 0 ( wth 20 order appromato for DKA method. Fgure 6(a shows plots of BesselJ 0 ( ad polyomal fucto for appromato. Fgure 6(b shows resolved roots. Horotal le the ceter of fgure correspods to real as. Meshes ths fgure show covergece traectores of resolved roots each terato step. We fd that 4 roots are sde ut crcle at least by umercal resoluto. Appromato Bessel -0-5 5 0 - Ut crcle r(209. Fg. 6 (a Plots of BesselJ 0 ( ad 20 order polyomal fucto for appromato, (b Numercal result I ths umercal treatmet, we used Mathematca, therefore we have doe symbolc treatmet for the computato wheever t s possble. Traectory

0-0 0 20-0 -0 0 20-0 -0 0 20-3.2 Hgher order appromato BesselJ 0 ( fucto ad ts 40 order appromato are show Fg. 7(a ad resoluto by DKA s show Fg. 7(b. We ca obta more roots at the same tme by the hgher order appromato method tha by the lower order oe. Ut crcle 0-0 0 20 coverg of the some rego usg polydscs. Each local polydsc overlap compactly wthout gaps the rego. Resoluto of the holomorphc fucto o each polydsc s possble wth proposed method the prevous secto. Small crcles ad tragles Fg. 9 correspod to roots by the method o each polydsc ad Fg.0 shows mage of cotuato of resoluto. Cosdet Error?, Neglect - r(406.7 Fg. 7 (a Plots of BesselJ0( ad 40 order appromato fucto, (b Numercal result 3.2 Cos( by 3rd order method wth 80 order appromatos Numercal resoluto of perodc fucto cos( ad ts appromato by 80 order polyomal fucto are show Fg. 8(a, ad ts roots by umercal resoluto wth 3rd order method are show Fg. 8(b. We ca fd may uecessary roots outsde ut crcle. It seems possble that the umber of these fcttous roots become smaller by ecellet seres epaso. By ths method, we obta the same umber of resolved roots to the umber of order of polyomal fucto for the appromato, whether we desre t or ot. Fg. 9 Cotuato or coecto wth polydscs for global resoluto of roots. r(5020 Appromato 4 3 2 0 0-0 0 20 30 - Ut crcle r(8030 Fg. 8 (a Plots of cos( by 80 order appromato fucto, (b resoluto by the method 4 Cotuato of local soluto We cosdered local resoluto of roots for the holomorphc fucto up to ths pot. I the followg secto, we cosder global dstrbuto of the roots by coecto ad cotuato. Fgure 9 llustrates global dstrbuto of roots by coecto wth may local resolutos by DKA. We cosder Fg. 0 Image of cotuato by proposed method. I ths case, we ca defe ad use covergece radus o each polydsc. Therefore, we wll obta relable roots ad fcttous roots by the method the each polydsc. Whe a root by the method a polydsc s ot foud aother polydsc where the rego of the both polydscs s overlapped, we must cosder such root s fcttous oe. Moreover, we must detfy or dstgush two close roots. The oe s obtaed a polydsc ad the other s obtaed aother polydsc. It s dffcult to detfy or dstgush such two roots overlapped rego, because both are very close ad we have o formato of the dstrbuto about the roots ths rego. As for the prescrpto of ths dffculty, we use square regos. The se of square regos are defed by the smallest radus of all polydscs the

rego. Sdes of each square rego cosst of two ds of boudares, whch are defed as Fg.. Closed r( Fg. Global cotuato by polydsc for the method. We oly allow resolved roots each square rego whch are covered by a polydsc that cludes the square wholly. Usg ths coecto, cotuato ad selecto, we ca obta all resolved roots wthout redudacy the whole rego. 5 Applcato of DKA method wth ths procedure We ca use above method to appromate u( (8. To costruct ut fucto u(, we treat as followg, (5 m m ( ( P r w ( ( ( (, Ope (,2,..., m, here w,,..,m are resolved roots by DKA or 3rd method sde polydsc. m (m s umber of roots sde polydsc of whch radus s (4. The we put, (6 / u ( f (/ P (, u ( u ( for the appromato of u(. Rough estmato of u( gves effectve formato for the eact treatmet of Weerstrass preparato theorem for f(, ad P ( each polydsc gves algebrac obect to represet orgal holomorphc fucto by dfferece (algebrac formula. 7 Summares ad Coclusos Eteded treatmet of DKA a d 3rd order method for holomorphc (aalytc fuctos are proposed. Weerstrass preparato theorem s treated from umercal pot of vew. Local dscrete represetatos of holomorphc fuctos are possble usg proposed eteded method. New umercal treatmet for evaluatg resdues of holomorphc fuctos wll be developed by ths approach. Refereces: [] W. C. Rheboldt, Methods for solvg Systems of Nolear Equatos, SIAM, 987. [2] T. Itoh, Algebrac Geometrcal Treatmet of Numercal Algorthms, WSEAS Trasactos o Mathematcs, Vol. 2, No., 2003, pp.30. [3] E. Durad, Soluto umérques des équatos algébrques, Masso, 960. [4] T. Yamamoto ad T. Ktagawa, Suchaseesyu, Scece-sha, Japa,99. [5] R. C. Gug ad H. Ross, Aalytc Fuctos of Several Varables, Pretce-Hal, Ic., Eglewood Clffs, N. J.,l965. Apped Holomorphc fucto C Defto : A comple-valued fucto f defed o C s called holomorphc fucto C f each pot w C has a ope eghborhood U, w U C, such that the fucto f has a power seres epresso v (A. v f( a ( w ( w v v 0 v v whch coverges for all U. Here C C C whch s Cartesa product of copes of comple plae, ad y are real umbers ad + y C, (,,. Defto 2: A ope polydsc C s a subset ( wr ; C of the form (A.2 ( wr ; ( w,, w; r,, r ; { C ; w < r, } the pot w C s called the ceter of the polydsc, ad r ( r,, r R,( r > 0 s called the polyradus. Note that polyomals the fuctos,..., are holomorphc all C. It s famlar result from elemetary aalyss that a power seres epaso of the form (A. s absolutely uformly coverget all sutable small ope polydscs ( wr ; cetered at the pot w. Moreover, the fucto f s holomorphc each varable separately throughout the doma whch t s aalytc.