Quadratic Bezier Homotopy Function for Solving System of Polynomial Equations

Similar documents
A new Approach for Solving Linear Ordinary Differential Equations

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

A Hybrid Variational Iteration Method for Blasius Equation

A MODIFIED METHOD FOR SOLVING SYSTEM OF NONLINEAR EQUATIONS

SOLVING NON-LINEAR SYSTEMS BY NEWTON s METHOD USING SPREADSHEET EXCEL Tay Kim Gaik Universiti Tun Hussein Onn Malaysia

On the Interval Zoro Symmetric Single-step Procedure for Simultaneous Finding of Polynomial Zeros

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b

Convexity preserving interpolation by splines of arbitrary degree

Inexact Newton Methods for Inverse Eigenvalue Problems

Fixed point method and its improvement for the system of Volterra-Fredholm integral equations of the second kind

The Quadratic Trigonometric Bézier Curve with Single Shape Parameter

DUE: WEDS FEB 21ST 2018

arxiv: v1 [math.co] 12 Sep 2014

Numerical Heat and Mass Transfer

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

Cubic Trigonometric B-Spline Applied to Linear Two-Point Boundary Value Problems of Order Two

EEE 241: Linear Systems

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

On the Multicriteria Integer Network Flow Problem

COMPOSITE BEAM WITH WEAK SHEAR CONNECTION SUBJECTED TO THERMAL LOAD

MMA and GCMMA two methods for nonlinear optimization

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law:

Chapter 4: Root Finding

NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS

Chapter 12. Ordinary Differential Equation Boundary Value (BV) Problems

Difference Equations

Application of B-Spline to Numerical Solution of a System of Singularly Perturbed Problems

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system

Formulas for the Determinant

Bézier curves. Michael S. Floater. September 10, These notes provide an introduction to Bézier curves. i=0

Review of Taylor Series. Read Section 1.2

Linear Feature Engineering 11

NUMERICAL DIFFERENTIATION

Beyond Zudilin s Conjectured q-analog of Schmidt s problem

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

Some modelling aspects for the Matlab implementation of MMA

On a direct solver for linear least squares problems

Speeding up Computation of Scalar Multiplication in Elliptic Curve Cryptosystem

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

Finite Element Modelling of truss/cable structures

ON A DETERMINATION OF THE INITIAL FUNCTIONS FROM THE OBSERVED VALUES OF THE BOUNDARY FUNCTIONS FOR THE SECOND-ORDER HYPERBOLIC EQUATION

MA 323 Geometric Modelling Course Notes: Day 13 Bezier Curves & Bernstein Polynomials

Nonlinear Overlapping Domain Decomposition Methods

The Analytical Solution of a System of Nonlinear Differential Equations

The Exact Formulation of the Inverse of the Tridiagonal Matrix for Solving the 1D Poisson Equation with the Finite Difference Method

Fuzzy Boundaries of Sample Selection Model

Global Sensitivity. Tuesday 20 th February, 2018

829. An adaptive method for inertia force identification in cantilever under moving mass

Kernel Methods and SVMs Extension

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS)

THE STURM-LIOUVILLE EIGENVALUE PROBLEM - A NUMERICAL SOLUTION USING THE CONTROL VOLUME METHOD

Binomial transforms of the modified k-fibonacci-like sequence

DETERMINATION OF TEMPERATURE DISTRIBUTION FOR ANNULAR FINS WITH TEMPERATURE DEPENDENT THERMAL CONDUCTIVITY BY HPM

Irregular vibrations in multi-mass discrete-continuous systems torsionally deformed

DEGREE REDUCTION OF BÉZIER CURVES USING CONSTRAINED CHEBYSHEV POLYNOMIALS OF THE SECOND KIND

Numerical Solutions of a Generalized Nth Order Boundary Value Problems Using Power Series Approximation Method

Report on Image warping

Bezier curves. Michael S. Floater. August 25, These notes provide an introduction to Bezier curves. i=0

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

The Study of Teaching-learning-based Optimization Algorithm

The Minimum Universal Cost Flow in an Infeasible Flow Network

New Method for Solving Poisson Equation. on Irregular Domains

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Lecture 12: Discrete Laplacian

COEFFICIENT DIAGRAM: A NOVEL TOOL IN POLYNOMIAL CONTROLLER DESIGN

Note 10. Modeling and Simulation of Dynamic Systems

Second Order Analysis

1 GSW Iterative Techniques for y = Ax

An efficient algorithm for multivariate Maclaurin Newton transformation

2. PROBLEM STATEMENT AND SOLUTION STRATEGIES. L q. Suppose that we have a structure with known geometry (b, h, and L) and material properties (EA).

Min Cut, Fast Cut, Polynomial Identities

On the Repeating Group Finding Problem

Significance of Dirichlet Series Solution for a Boundary Value Problem

Support Vector Machines. Vibhav Gogate The University of Texas at dallas

Computing Correlated Equilibria in Multi-Player Games

Chapter Newton s Method

Chapter 3 Differentiation and Integration

Curve Fitting with the Least Square Method

General viscosity iterative method for a sequence of quasi-nonexpansive mappings

More metrics on cartesian products

A NUMERICAL COMPARISON OF LANGRANGE AND KANE S METHODS OF AN ARM SEGMENT

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

The KMO Method for Solving Non-homogenous, m th Order Differential Equations

Lecture 13 APPROXIMATION OF SECOMD ORDER DERIVATIVES

Solving Fractional Nonlinear Fredholm Integro-differential Equations via Hybrid of Rationalized Haar Functions

Module 2. Random Processes. Version 2 ECE IIT, Kharagpur

International Conference on Advanced Computer Science and Electronics Information (ICACSEI 2013) equation. E. M. E. Zayed and S. A.

Research Article Cubic B-Spline Collocation Method for One-Dimensional Heat and Advection-Diffusion Equations

Relaxation Methods for Iterative Solution to Linear Systems of Equations

Appendix B. The Finite Difference Scheme

MATH 829: Introduction to Data Mining and Analysis The EM algorithm (part 2)

Interactive Bi-Level Multi-Objective Integer. Non-linear Programming Problem

Joint Statistical Meetings - Biopharmaceutical Section

Generalized Linear Methods

10.34 Numerical Methods Applied to Chemical Engineering Fall Homework #3: Systems of Nonlinear Equations and Optimization

Modelli Clamfim Equazione del Calore Lezione ottobre 2014

Transcription:

MATEMATIKA, 2013, Volume 29, Number 2, 159 171 c Department of Mathematcal Scences, UTM Quadratc Bezer Homotopy Functon for Solvng System of Polynomal Equatons 1 Hafzudn Mohamad Nor, 2 Ahmad Izan Md. Ismal and 3 Ahmad Abdul Majd 1,2,3 School of Mathematcal Scences, Unverst Sans Malaysa, 11800 Gelugor, Penang, Malaysa e-mal: 1 hafz97my2001@yahoo.com, 2 zan@cs.usm.my, 3 majd@cs.usm.my Abstract We compare standard homotopy functon wth a proposed quadratc Bezer homotopy functon to see whch method has greater applcablty and greater accuracy. We test the methods on system of polynomal equatons by usng Newton-Homotopy Contnuaton method. The results obtaned ndcate the superor accuracy of our proposed quadratc Bezer homotopy functon. Keywords Numercal method; Polynomal equatons; Homotopy functon. 2010 Mathematcs Subject Classfcaton 65H05; 65H10; 65H20. 1 Introducton In ths paper we consder the soluton of the system of n polynomal equatons f 1 (x 1, x 2,, x n ) f 2 (x 1, x 2,, x n ) F(x) =. = 0 (1) f n (x 1, x 2,, x n ) where x = {x 1, x 2,..., x n } usng homotopy contnuaton method (HCM). A popular method for solvng (1) s the Newton method whch s derved from the Taylor expanson seres [1]. In recent years, the study of the soluton of polynomal equatons especally on system of polynomal equatons usng methods based on the concept of homotopy from Topology has attracted consderable nterest. Rafq and Awas [2] stated that the homotopy contnuaton method (HCM) has been known snce the 1930s. However, more recent work n the development of homotopy concepts for nonlnear algebrac equaton was developed n the 1970s [3]. Ths was followed by works of Garca and Zangwll [4], Melhem and Rhenboldt [5], Morgan [6] and Watson [7,8]. Alexander and Yorke [3] stated that HCM nvolves numercally fndng the soluton of a problem by startng the soluton from the soluton of a known problem and contnung the soluton as the known problem s homotoped to the gven problem. Alexander and Yorke [3] also descrbed the connecton between algebrac topology and the contnuaton method. Garca and Zangwll [4] suggested a procedure to obtan all solutons to certan systems of n equatons n the complex doman. Let z q + P (z) = 0, = 1, 2,...,n. (2) where z s the complex vector, q s the large nteger and P(z) = F(z) 1. By rewrtng (2), we obtan (z q 1) + (P (z) + 1) = 0 (3)

160 Hafzudn Mohamad Nor, Ahmad Izan Md. Ismal and Ahmad Abdul Majd To start wth, the authors defned homotopy functon as (z q 1) + t(p (z) + 1) = 0 (4) where z s the complex vector, q s the large nteger and t [0, 1] s a load parameter. Snce 1 0, thus we wll get a trval soluton when t = 0. Then, the authors rewrte (4) as z q By assgnng z q From (6), we have As t 1, we have (1 t)(z q 1) + t(p (z) + 1) = 0 (5) 1 = G(z) and P (z) + 1 = F(z), then we have (1 t)g(z) + tf(z) = 0 (6) F(z) G(z) = 1 1 t (7) F(z) lm t 1 G(z) = 0 (8) Eq. (8) showed the relatonshp between the auxlary homotopy functon and the gven functon n the homotopy functon (6). Garca and Zangwll [4] also dscussed the meanng of the terms - homotopy paths, monotoncty of the paths, smplcal pvotng and pecewse lnear paths. Melhem and Rhenboldt [5] compared several methods from local teratve and contnuaton path methods for determnng turnng ponts. For nstance, Abbott s method [9], Moore and Spence s method [10], Seydel s method [11], Smpson s method [12], Rhennboldt s method [13] and so on. The results showed the Rhenboldt s method s most approprate because t has the hghest degree of relablty among all methods [5]. Snce HCM s a method of solvng dvergence problem, therefore a body of knowledge about how to determne the turnng ponts are become mportant. Morgan [6] used SYMPOL (systems of polynomals) to conduct the mplementaton. SYMPOL was desgned to fnd all solutons to a system of n polynomal equatons wth complex coeffcents n n unknowns. The author also nvestgated three cases of choosng auxlary homotopy functon G(x) havng a theorem developed n [6]. The auxlary functons selected to form the homotopy functons were () H(x, t) = (1 t)(x x 0 ) + t( x 2 + 1) (9) () H(x, t) = (1 t)(x 2 x 2 0) + t( x 2 + 1) (10) () H(x, t) = (1 t)(x 3 x 3 0 ) + t( x2 + 1) (11) where the target functon was F(x) = x 2 + 1 = 0. Morgan [6] also descrbed the basc concepts of HCM such as homotopy functon and the auxlary homotopy functon. Homotopy functon, denoted as H(x, t), s a connecton between the start and target functons, whle the auxlary homotopy functon s the startng functon. Accordng to Watson [7], homotopy methods are theoretcally powerful, and f constructed and mplemented properly, are robust, numercally stable, accurate, and practcal.

Quadratc Bezer Homotopy Functon for Solvng System of Polynomal Equatons 161 In [8], the homotopy functon was defned n the complex doman. The choce of auxlary homotopy functon s defned by G(x) = b x d a, = 1, 2,...,n (12) where a and b are nonzero complex numbers. There are two types of total degree, denoted by d and d. The total degree of polynomal F(x), d s d = max k n d jk (13) j=1 and the total degree of the entre system (1), d s d = d 1 d 2... d n. (14) Accordng to Kotsreas [14], the total number of geometrcally solated solutons and solutons at nfnty s no more than that gven by equaton (14). By usng (12) and (13), the nvestgated homotopy functons by Morgan [6] can be smplfed as () H(x, t) = (1 t)(x x 0 ) + tf(x) (15) () H(x, t) = (1 t)(x d x d 0 ) + tf(x) (16) () H(x, t) = (1 t)(x d+1 x d+1 0 ) + tf(x). (17) Jalal and Seader [15] analyzed the stablty of multphase and reactng systems by usng HCM. The accuracy of the ntal guess was not mportant. Jalal and Seader focused on the use of the Newton, fxed-pont, and affne homotopes whch are as follows () Newton homotopy () Fxed-pont homotopy H(x, t) = tf(x) + (1 t)[f(x) F(x 0 )] (18) () Affne homotopy H(x, t) = tf(x) + (1 t)(x x 0 ) (19) H(x, t) = tf(x) + (1 t)f (x 0 )(x x 0 ) (20) where F(x) s the polynomal equatons and t [0, 1]. Grtton et al. [16] classfed HCM as a global method whch the user can use to fnd the soluton from an arbtrary ntal guess. For local methods such as the Newton method, the user should have suffcent knowledge regardng the locaton of a root to determne the ntal guess. The closer the ntal guess s to the soluton, the more effcent s the local method. Otherwse, the teratve scheme of the local method wll dverge away from the

162 Hafzudn Mohamad Nor, Ahmad Izan Md. Ismal and Ahmad Abdul Majd actual soluton. Global methods overcome the problem of choosng the approprate ntal guess. Grtton et al. [16] used HCM to study 16 chemcal engneerng problems nvolvng sothermal flash, knetcs n a strred reactor, azeotropc-pont, flow n a smooth ppe, chemcal equlbrum and others. One problem consdered nvolved the equaton, F(x) = 8(4 x)2 x 2 0.186 = 0 (21) (6 3x) 2 (2 x) where x s the fractonal converson of ntrogen. The solutons were tracked by usng HCM and adjustng auxlary functons (.e. the ntal guess). Allgower and Georg [17] defned the polynomal systems (1) as n 1 F(z) = z n + a j z j (22) and equaton (22) s a monc polynomal n whch the coeffcent of z n s equal to 1. Lettng G(z) = z n + b 0 where b 0 0, the authors defned homotopy functon as Hence, we have j=0 H(z, t) = (1 t)g(z) + tf(z). (23) n 1 H(z, t) = z n + t a j z j + (1 t)b 0 + ta 0 (24) j=0 where H(z, t) s measured n complex doman H : C [0, 1] C. Wu [18-22] has conducted extensve research on HCM. Wu [18] ntroduced Ancent Chnese Homotopy method (ACHM) and Wu [19] studed the convergence of Newton-Homotopy contnuaton method (NHCM). Wu [20] compared the tradtonal Adoman Decomposton method wth the Adoman-homotopy contnuaton method (AHCM). Further, Wu [21] nvestgated the crtera of adjustable auxlary homotopy functon and ts gudelnes and Wu [22] developed secant-homotopy contnuaton method (SHCM) from tradtonal secant method. Palancz et al. [23] descrbed HCM n a smpler way. HCM was defned as a method that deforms contnuously from the known roots of the start system nto the roots of the target system. Contnuaton graphs were used to llustrate the mathematcal problems. A smple equaton was used to demonstrate the homotopy contnuaton concepts. Then, Palancz et al. [23] extended the concepts by dscussng nonlnear geodetc problems such as resecton, GPS postonng, as well as affne transformaton. Rahman et al. [24] concerned wth the use of homotopy functons H(x, t) = tf(x) + (1 t)[(x x 0 ) + (F(x) F(x 0 )] (25) to track the approxmate solutons. The authors chose G(x) as a lnear combnaton of fxed pont and Newton functons. In most of above research, the focus was on the auxlary homotopy functon as well as homotopy contnuaton method and not the homotopy functon. The authors [1-2,4,6,14-24] were more comfortable to use standard homotopy functon rather than others. Ths paper wll ntroduce a new homotopy functon whch wll be called the Quadratc Bezer Homotopy Functon (QBHF).

Quadratc Bezer Homotopy Functon for Solvng System of Polynomal Equatons 163 2 Standard Homotopy Functon All homotopy functon H(x, t) mentoned on above used the standard homotopy functon. Let us consder the standard homotopy functon where H(x, t) = (1 t)g(x) + tf(x) (26) G(x) = F(x) = g 1 (x 1, x 2,, x n ) g 2 (x 1, x 2,, x n ). g n (x 1, x 2,, x n ) f 1 (x 1, x 2,, x n ) f 2 (x 1, x 2,, x n ). f n (x 1, x 2,, x n ) and t s an arbtrary parameter whch can vary from 0 to 1,.e. t [0, 1]. Thus, we wll have H(x, 0) = G(x). (29) (27) (28) H(x, 1) = F(x). (30) The homotopy functon cannot stand alone; t must be followed by a method whch s called homotopy contnuaton method. Accordng to Burden and Fares [1], the formula of Newton HCM s as follows where x +1 = x [D x H(x, t)] 1 H(x, t), = 1, 2,..., k. (31) D x H(x, t) =. H 1 H 1 x 1 H 2 H 2 x 1 H n x 1 x 2 x 2....... H n x 2 D x H(x, t) s called the Jacoban matrx. To facltate better understandng, let us consder the 2-dmensonal of homotopy functon whch s n = 1. H(x, t) = (1 t)g(x) + tf(x) (32) where g(x) s an auxlary homotopy functon and f(x) s a gven functon whch s scalar. H 1 x n H 2 x n H n x n Example 2.1. Consder the followng smple equaton defned by Kotsreas [14]. f(x) = (x 2 1 4 )(x2 4) = 0 (33) and the auxlary homotopy functon s g(x) = x 2 1. So that, the homotopy functon (30) wll be H(x, t) = (1 t)(x 2 1) + t(x 2 1 4 )(x2 4) (34)

164 Hafzudn Mohamad Nor, Ahmad Izan Md. Ismal and Ahmad Abdul Majd Snce we need to solve H(x, t) = 0, there are a set equatons of H (x, t /k ) = 0 and there are a set solutons for x where = 0, 1, 2,, k and k = number of teratons. If there are k teratons, the number of equaton nvolved s k+1. Let k =10, so that the equaton vares from 0 untl 1. Graphcally, t can be represented as Fgure 1. Fgure 1: Homotopy Path of Equaton (34) Fgure 1 shows the movement of H(x 0, 0) = x 2 1 = g(x) to H(x 10, 1) = (x 2 1 4 )(x2 4) = f(x) wth t beng unformly ncreased by 0.1 n equaton (34). x 0 = 1 wll move to the x 10 = 2.000618931 and x 10 = -2.000618931 when x 0 = 1. Both approxmatons wll have the same value of f(x 10 ).e. f(x 10 ) = 9.29 10 3 (35) However, ths approxmaton can be mproved by usng a new homotopy functon whch wll be called the Quadratc Bezer Homotopy Functon. Ths wll be dscussed n the next secton. 3 Quadratc Bezer Homotopy Functon Suppose that we want to solve followng a system of n polynomal equatons F(x) = 0. (36) where x = {x 1, x 2, x 3,, x n 1, x n }. We ntroduce a new homotopy functon vz H 2 (x, t) = (1 t) 2 G(x) + 2t(1 t)[(1 t)g(x) + tf(x)] + t 2 F(x). (37) Ths new homotopy functon (37) can also be wrtten as H 2 (x, t) = (1 t) 2 G(x) + 2t(1 t)h(x, t) + t 2 F(x) (38)

Quadratc Bezer Homotopy Functon for Solvng System of Polynomal Equatons 165 where H(x, t) s the standard homotopy functon. Ths new homotopy functon fulflls the two boundary condtons.e. (29) and (30) are stll satsfed when we substtute t = 0 and t = 1 respectvely nto (38). It s nterestng to note that there s a smlar functon of homotopy functon between (26) and (38) when t = 1 2 where H(x, 1 2 ) = H 2(x, 1 2 ) = 1 2 G(x) + 1 2 F(x). (39) Snce we want to solve F(x) = 0, therefore H 2 (x, t) s set to zero by varyng the parameter t from 0 to 1. In other words, we start from auxlary functon, G(x 0 ) = 0 and fnsh when F( x) = 0. As dscussed before, the soluton moves from x 0 untl x and the curves wll move from H 2 (x 0, 0) untl H 2 ( x, 1). The dea of ths new homotopy functon comes from De Casteljau Algorthm [25]. It s known that, De Casteljau Algorthm descrbes the movement of pont n a curve. Homotopy s a movement of a curve to another curve [14,23]. Therefore, we beleve that there s relaton between the De Casteljau algorthm and homotopy concepts. In short, lnear and quadratc curves for De Casteljau and homotopy can be formed as n Table 1. Table 1: Lnear and Quadratc for De Casteljau and Homotopy De Casteljau, P(t) Homotopy, H(x, t) Lnear (1 t)p 0 + tp 1 (1 t)g(x) + tf(x) Quadratc (1 t) 2 P 0 + 2t(1 t)p 1 + t 2 P 2 (1 t) 2 G(x)+ 2t(1 t)[(1 t)g(x) + tf(x)] + t 2 F(x). Let s observe how we obtan (38) wth three reference curves. The recursve constructon of quadratc Bezer homotopy functon s llustrated n Fgure 2. We note that A(x, t) = (1 t)g(x) + th(x, t), B(x, t) = (1 t)h(x, t) + tf(x) Therefore H 2 (x, t) = (1 t)a + tb (40) Then, we have H 2 (x, t) = (1 t)[(1 t)g(x) + th(x, t)] + t [(1 t)h(x, t) + tf(x)] = (1 t) 2 G(x) + 2t(1 t)h(x, t) + t 2 f(x) = H 0 B0(t) 2 + H 1 B1(t) 2 + H 2 B2(t) 2 2 = H B 2 (t) (41) =0

166 Hafzudn Mohamad Nor, Ahmad Izan Md. Ismal and Ahmad Abdul Majd Fgure 2: Recursve Constructon of Quadratc Bezer Homotopy Functon where H 0 = G(x) for t = 0 H 1 = H(x, t) for t (0, 1) H 2 = F(x) for t = 1. B 2 (t) s a Bernsten functon whch s defned as [25] ( ) B 2 2 (t) = (1 ) 2 t 2! =! (2 )! (1 )2 t (42) where : 0, 1, 2, t [0, 1]. An nterestng property of Bernsten functon for standard homotopy functon and QBHF s that the summaton s equal to one. Lnear : Quadratc : 1 B 1(t) =0 = (1 t) + t = 1, 1 B 2(t) = (1 t)2 + 2t(1 t) + t 2 =0 = 1. (43) (44) Both (43) and (44) show that the sum of bnomal expanson always fulfll one of the convex hull property [25]. Another convex hull property s B 2 (t) 0, : 0, 1, 2, t [0, 1]. (45)

Quadratc Bezer Homotopy Functon for Solvng System of Polynomal Equatons 167 Ths contnuaton technque cannot stand alone n that t must be combned wth other methods such as Newton, secant, Adoman method and so on. By these combnatons, the name of the newly developed method wll then change to Newton-Homotopy, Secant- Homotopy, Adoman-Homotopy and so on. Now, we wll consder several examples that compares the standard and new homotopy functons. The method chosen s Newton- Homotopy contnuaton method (NHCM). The formula of classcal Newton method s well-known. The formula of NHCM s x +1 = x [D x H 2 (x, t)] 1 H 2 (x, t), = 1, 2,...,k. (46) where H 2 (x, t) s Quadratc Bezer homotopy functon as (38). 4 Numercal Experments and Dscusson Example 4.1. Consder the followng system of equatons [2] : f 1 (x, y) = x 2 2x y + 1 2 = 0, f 2 (x, y) = x 2 + 4y 2 4 = 0. (47) The auxlary homotopy functon s g 1 (x) = x, g 2 (y) = y and the ntal value (x 0, y 0 ) = (0, 0)are used for NHCM. The results are shown n Table 2 by varyng the number of teratons. Table 2: Comparson between Standard and New Homotopy Functons for Equaton (47) Number of teratons Standard Homotopy Functon 10 f 1 = 4.90 10 3 f 2 = 2.93 10 2 100 f 1 = 3.83 10 5 f 2 = 2.96 10 4 1000 f 1 = 3.72 10 7 f 2 = 2.96 10 6 Quadratc Bezer Homotopy Functon CPU Tme, second f 1 = 2.38 10 4 f 2 = 2.63 10 3 0.0070004 f 1 = 3.28 10 8 f 2 = 2.63 10 7 0.0300017 f 1 = 3.33 10 12 f 2 = 2.66 10 11 0.2180124 where ( x, ỹ) = (1.90067672637127,0.31121856542355) Example 4.2. Consder the followng example [26]: f 1 (x, y, z) = x 2 + y 2 + z 2 1 = 0 f 2 (x, y, z) = 2x 2 + y 2 4z = 0 f 3 (x, y, z) = 3x 2 4y 2 + z 2 = 0 (48) The auxlary homotopy functon, g 1 (x) = x, g 2 (y) = y and g 3 (z) = z and ntal value (x 0, y 0, z 0 ) = (0, 0, 0) are used for NHCM. The results are shown n Table 3.

168 Hafzudn Mohamad Nor, Ahmad Izan Md. Ismal and Ahmad Abdul Majd Table 3: Comparson between Standard and New Homotopy Functons for Equaton (48) Number of teratons Standard Homotopy Functon 10 f 1 = 2.12 10 3 f 2 = 1.21 10 3 f 3 = 1.75 10 3 100 f 1 = 1.82 10 5 f 2 = 1.04 10 5 f 3 = 1.56 10 5 1000 f 1 = 1.79 10 7 f 2 = 1.02 10 7 f 3 = 1.54 10 7 Quadratc Bezer Homotopy Functon f 1 = 1.36 10 4 f 2 = 7.23 10 5 f 3 = 1.13 10 4 f 1 = 1.59 10 8 f 2 = 9.02 10 9 f 3 = 1.36 10 8 f 1 = 1.61 10 12 f 2 = 9.14 10 13 f 3 = 1.38 10 12 where ( x, ỹ, z) = (0.69828860997219,-0.62852429796055,0.34256418968992) CPU Tme, second 0.0200011 0.1260072 1.1730670 Example 4.3. Consder the followng example [6]: f 1 (w, x, y, z) = x + 10y = 0 f 2 (w, x, y, z) = 5(z w) = 0 f 3 (w, x, y, z) = (y 2z) 2 = 0 f 4 (w, x, y, z) = 10(x w) 2 = 0 (49) The auxlary homotopy functon, g 1 (w) = w 1, g 2 (x) = x 4, g 3 (y) = y 1, g 4 (z) = z 2 and ntal value s (w 0, x 0, y 0, z 0 ) = (1, 4, 1, 2) are used. The results are shown n Table 4. The results n Table of Eq. (47), (48), and (49) show that Quadratc Bezer Homotopy Functon has better performance than the standard homotopy functon for solvng the consdered system of nonlnear algebrac equaton. Ths has been ascertaned by usng the followng stoppng crtera F( X k+1 ) < ε (50) where ε = 10 20. CPU tme s assumed not to be an mportant consderaton. 5 Concluson From the partcular set of examples chosen, the use of Quadratc Bezer Homotopy Functon (QBHF) s better than the use of standard homotopy functon. It should be noted though that for the partcular set of examples consdered, the accuracy of approxmate solutons ncreases when the numbers of teratons ncrease. Acknowledgments Part of ths research has been supported by a scholarshp from the School of Mathematcal Scences USM and Kementeran Pengajan Tngg Malaysa.

Quadratc Bezer Homotopy Functon for Solvng System of Polynomal Equatons 169 Table 4: Comparson between Standard and New Homotopy Functons for Equaton (49) Number of teratons Standard Homotopy Functon 10 f 1 = 2.78 10 17 f 2 = 7.14 10 16 f 3 = 2.93 10 2 f 4 = 7.21 10 2 100 f 1 = 6.37 10 16 f 2 = 2.72 10 15 f 3 = 2.88 10 3 f 4 = 5.61 10 3 1000 f 1 = 6.67 10 16 f 2 = 2.67 10 15 f 3 = 2.86 10 4 f 4 = 5.78 10 4 Quadratc Bezer Homotopy Functon f 1 = 3.89 10 16 f 2 = 4.97 10 16 f 3 = 1.79 10 2 f 4 = 2.27 10 2 f 1 = 6.14 10 16 f 2 = 4.09 10 16 f 3 = 1.32 10 4 f 4 = 2.60 10 4 f 1 = 4.20 10 16 f 2 = 4.00 10 16 f 3 = 1.33 10 6 f 4 = 2.65 10 6 CPU Tme, second 0.1240071 0.4270244 3.4601979 where ( w, x, ỹ, z) = (0.00050521152754,0.00142117446223,-0.00014211744622,0.00050521152754) and the actual soluton s (w, x,y, z) = (0,0,0,0) as stated n Morgan [6] References [1] R. L. Burden, J. D. Fares, Numercal Analyss, 9 th Internatonal Edton, Brooks/Cole, Cencag Learnng. 2011. [2] Rafq, A. and Awas, M. Convergence on the homotopy contnuaton method. Int. J. of Appl. Math. And Mech. 2008. 4(6) :62 70. [3] Alexander, J. C. and Yorke, J. A. The homotopy contnuaton method: numercally mplementable topologcal procedures, Trans. Amer. Math. Soc. 1978. 242: 271 284. [4] Garca, C. B. and Zangwll, W. I. Fndng all solutons to polynomal systems and other systems of equatons, Mathematcal Programmng. 1979. 16 :159 176. [5] Melhem, R. C. and Rhenboldt, W. C. A comparson of methods for determnng turnng ponts of nonlnear equatons, Computng. 1982. 29 : 201 226. [6] Morgan, A. P. A method for computng all solutons to systems of polynomal equatons, ACM Transacton on Mathematcal Software. 1983. 9(1) :1-17. [7] Watson, L. T. Numercal lnear algebra aspects of globally convergent homotopy methods, Techncal Report. 1986. [8] Watson, L. T. Globally convergent homotopy methods: A tutoral, Elsever Scence Publshng Co. Inc. 1989: 369 396.

170 Hafzudn Mohamad Nor, Ahmad Izan Md. Ismal and Ahmad Abdul Majd [9] Abbott, J. P. An effcent algorthm for the determnaton of certan bfurcaton ponts, J. of Computer and Appled Mathematcs 1978. 4 : 19 27. [10] Moore, G. and Spence, A. The calculaton of turnng ponts of nonlnear equatons, SIAM J. Number Analyss.1980.17 : 567 576. [11] Seydel, R. Numercal computaton of branch ponts n nonlnear equatons, Numercal Math. 1979. 33 : 339 352. [12] Smpson, R. B., Smpson method for numercal determnaton of bfurcaton states of nonlnear systems of equatons, SIAM Journal Numercal Analyss. 1975.12 : 439 451. [13] Rhenboldt, W. C. Computaton of crtcal boundares on equlbrum manfolds, Unversty of Pttsburgh, Insttute for Computatonal Mathematcs and Appled, Techncal Report, ICMA-80-20, SIAM Journal Numercal Analyss.1982.19: 653 669. [14] Kotsreas, I. S. Homotopes and polynomal system solvng I: Basc Prncples. ACM- SIGSAM Bulletn. 2001. 35: 19 32. [15] Jalal, F. and Seader,J. D. Use of homotopy-contnuaton method n stablty analyss of multphase reactng systems, Computers and Chemcal Engneerng. 2000. 24: 1997 2000. [16] Grtton, K. S., Seader, J. D. and Ln,W. Global homotopy contnuaton procedures for seekng all roots of a nonlnear equaton, Computers and Chemcal Engneerng. 2001. 25: 1003 1019. [17] Allgower, E. L. and Georg, K. Introducton to numercal contnuaton methods, Socety of Industral and Appled Mathematcs. 2003: 132 135. [18] Wu, T. M. A modfed formula of ancent Chnese algorthm by the homotopy contnuaton technque, Appled Mathematcs and Computaton. 2005. 165 : 31-35. [19] Wu, T. M. A study of convergence on the Newton-Homotopy contnuaton method. Appled Mathematcs and Computaton.2005.168: 1169-1174. [20] Wu, T. M. A new formula of solvng nonlnear equatons by Adoman and homotopy methods. Appled Mathematcs and Computaton. 2006.172: 903-907. [21] Wu, T. M. Solvng nonlnear equatons by the Newton Homotopy contnuaton method wth adjustable auxlary homotopy functon, Appled Mathematcs and Computaton. 2006. 173 : 383-388. [22] Wu, T. M. The Secant-Homotopy contnuaton method. Chaos Soltons and Fractals. 2007. 32 : 888-892. [23] Palancz, B., Awange, J. L., Zaletnyk, P. and Lews, R. H. Lnear homotopy soluton of nonlnear systems of equatons n geodesy, Sprnger-Verlag.2010 :79-95.

Quadratc Bezer Homotopy Functon for Solvng System of Polynomal Equatons 171 [24] Rahman, S. K., Jalal, F., Seader, J. D. and Whte, R. E. A new homotopy for seekng all real roots of a nonlnear equaton, Computers and Chemcal Engneerng. 2011.35: 403 411. [25] Agoston, M. K. Computer Graphc and Computer Modelng, 3 rd Edton, Sprnger- Verlag,Cupertno CA 95014, 2004. [26] Noor, M. A. and Waseem, M. Some teratve methods for solvng a system of nonlnear equatons, Computers and Mathematcs wth applcatons. 2009. 57: 101-106.