TR/95 February Splines G. H. BEHFOROOZ* & N. PAPAMICHAEL

Similar documents
TR/01/89 February An O(h 6 ) cubic spline interpolating procedure for harmonic functions. N. Papamichael and Maria Joana Soares*

The Finite Element Method: A Short Introduction

Stanford University CS254: Computational Complexity Notes 7 Luca Trevisan January 29, Notes for Lecture 7

Convexity preserving interpolation by splines of arbitrary degree

Solution for singularly perturbed problems via cubic spline in tension

APPENDIX A Some Linear Algebra

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0

TR/28. OCTOBER CUBIC SPLINE INTERPOLATION OF HARMONIC FUNCTIONS BY N. PAPAMICHAEL and J.R. WHITEMAN.

Numerical Simulation of One-Dimensional Wave Equation by Non-Polynomial Quintic Spline

5 The Laplace Equation in a convex polygon

Numerical Heat and Mass Transfer

Lecture 2: Numerical Methods for Differentiations and Integrations

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

On Pfaff s solution of the Pfaff problem

Lecture 13 APPROXIMATION OF SECOMD ORDER DERIVATIVES

A Discrete Approach to Continuous Second-Order Boundary Value Problems via Monotone Iterative Techniques

A new Approach for Solving Linear Ordinary Differential Equations

Formulas for the Determinant

Randić Energy and Randić Estrada Index of a Graph

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

NUMERICAL DIFFERENTIATION

ALGORITHM FOR THE CALCULATION OF THE TWO VARIABLES CUBIC SPLINE FUNCTION

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Foundations of Arithmetic

Maejo International Journal of Science and Technology

( ) [ ( k) ( k) ( x) ( ) ( ) ( ) [ ] ξ [ ] [ ] [ ] ( )( ) i ( ) ( )( ) 2! ( ) = ( ) 3 Interpolation. Polynomial Approximation.

Competitive Experimentation and Private Information

Nice plotting of proteins II

8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS

Multivariate Ratio Estimator of the Population Total under Stratified Random Sampling

One-sided finite-difference approximations suitable for use with Richardson extrapolation

On a nonlinear compactness lemma in L p (0, T ; B).

Positivity Preserving Interpolation by Using

More metrics on cartesian products

On Finite Rank Perturbation of Diagonalizable Operators

A Spline based computational simulations for solving selfadjoint singularly perturbed two-point boundary value problems

Lecture 6/7 (February 10/12, 2014) DIRAC EQUATION. The non-relativistic Schrödinger equation was obtained by noting that the Hamiltonian 2

The Order Relation and Trace Inequalities for. Hermitian Operators

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

Solving Singularly Perturbed Differential Difference Equations via Fitted Method

Lecture 5 Decoding Binary BCH Codes

Adaptive Kernel Estimation of the Conditional Quantiles

COMP4630: λ-calculus

Least squares cubic splines without B-splines S.K. Lucas

Causal Diamonds. M. Aghili, L. Bombelli, B. Pilgrim

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Not-for-Publication Appendix to Optimal Asymptotic Least Aquares Estimation in a Singular Set-up

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

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family

The finite element method explicit scheme for a solution of one problem of surface and ground water combined movement

Bernoulli Numbers and Polynomials

Two point fuzzy boundary value problem with eigenvalue parameter contained in the boundary condition

5 The Rational Canonical Form

Lecture 12: Discrete Laplacian

EEE 241: Linear Systems

Perron Vectors of an Irreducible Nonnegative Interval Matrix

Iranian Journal of Mathematical Chemistry, Vol. 5, No.2, November 2014, pp

1 Introduction We consider a class of singularly perturbed two point singular boundary value problems of the form: k x with boundary conditions

A CHARACTERIZATION OF ADDITIVE DERIVATIONS ON VON NEUMANN ALGEBRAS

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

Determinants Containing Powers of Generalized Fibonacci Numbers

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

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

The optimal delay of the second test is therefore approximately 210 hours earlier than =2.

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction

CHAPTER 14 GENERAL PERTURBATION THEORY

SL n (F ) Equals its Own Derived Group

Shuai Dong. Isaac Newton. Gottfried Leibniz

Linear Approximation with Regularization and Moving Least Squares

An efficient algorithm for multivariate Maclaurin Newton transformation

Applied Mathematics Letters

Module 3: Element Properties Lecture 1: Natural Coordinates

Problem Set 4: Sketch of Solutions

Exercise Solutions to Real Analysis

The exponential map of GL(N)

Finite Differences, Interpolation, and Numerical Differentiation

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

Supplemental document

Report on Image warping

Errors for Linear Systems

New Method for Solving Poisson Equation. on Irregular Domains

Polynomial Regression Models

Composite Hypotheses testing

ELASTIC WAVE PROPAGATION IN A CONTINUOUS MEDIUM

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

Initial-value Technique For Singularly Perturbed Two Point Boundary Value Problems Via Cubic Spline

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space.

Group Analysis of Ordinary Differential Equations of the Order n>2

Difference Equations

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS

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

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

Formal solvers of the RT equation

Monotonic Interpolating Curves by Using Rational. Cubic Ball Interpolation

STAT 3008 Applied Regression Analysis

arxiv: v1 [math.co] 12 Sep 2014

Asymptotics of the Solution of a Boundary Value. Problem for One-Characteristic Differential. Equation Degenerating into a Parabolic Equation

Beyond Zudilin s Conjectured q-analog of Schmidt s problem

Transcription:

TR/9 February 980 End Condtons for Interpolatory Quntc Splnes by G. H. BEHFOROOZ* & N. PAPAMICHAEL *Present address: Dept of Matematcs Unversty of Tabrz Tabrz Iran.

W9609

A B S T R A C T Accurate end condtons are derved for quntc splne nterpolaton at equally spaced nots. Tese condtons are n terms of avalable functon values at te nots and lead to 0( 6 ) covergence unformly on te nterval of nterpolaton.

- -. Introducton Let Q be a quntc splne on [ab] wt equally spaced nots x a ; 0l... (. ) were (b-a)/. Ten Q C 4 [ab] and n eac of te ntervals [x - x ] ;... Q s a quntc polynomal. Te set of all suc quntc splnes forms a lnear space of dmenson wc we denote by Sp(). Gven te set of values y ; -0l... were y y ) y c n [ab] n 6 we consder te problem of constructng an nterpolatory Q Sp() suc tat Q ) y ; 0.... (.) Snce dm Sp() te nterpolaton condtons (.) are not suffcent to determne Q unquely and four addtonal lnearly ndependent condtons are always needed for ts purpose. Tese are usually taen to be end condtons.e. condtons mposed on Q or ts dervatves Q (J) ; 4 near te two end ponts a and b. As mgt be expected te coce of end condtons plays a crtcal role on te qualty of te splne approxmaton. It s well nown tat te best order of approxmaton wc can be aceved by an nterpolatory quntc splne Q s Q - y 0( 6 ) were. denotes te unform norm on [ab]. Suc order of convergence s obtaned f for example te end condtons Q ) y Q ( x ) y ; 0

- - are used. However tese condtons requre nowledge of te frst dervatve of y at te four ponts x.x - ; 0 and n an nterpolaton problem ts nformaton s not usually avalable. Te natural end condtons Q (r) (a) Q (r) (b) 0 ; r 4 do not requre any addtonal nformaton but te resultng natural quntc splne does not ave 0( 6 ) convergence unformly on [ab]. Te purpose of te present paper s to derve end condtons for quntc splne nterpolaton at equally spaced nots wc depend only on te gven functon values at te nots and lead to 0 ( 6 ) convergence unformly on [ab]. We derve a class of suc end condtons n Secton by generalzng te cubc splne results of Beforooz and Papamcael (979) to te case of quntc splne nterpolaton.. Prelmnary Results To smplfy te presentaton we use trougout te abbrevatons () () m Q ) M Q ) n Q ) and N (4) Q ); I 0.. (.) Te followng quntc splne denttes are needed for te analyss of Secton : m - 6m - 66m 6m m {-y - - 0y - 0y y } ;.... - (.) M - 6M - 66M 6M M 0 { y - y - -6y y y };... - (.)

- - n - 6n - 66n 6n n 60 {-y - y - -y y } ;... - (.4) N - 6N -l 66N 6N l N 0 4 { y - - 4y - 6y - 4y y } ;... - (.) N - 4N - N - 6 { M _ - M M } 0;... - (.6) 60 { m - m m } - {n - 4n n } 0 {y -y - } ;... - l (.7) 8 {m - m - } - {M - - 6M M } 0 {y _ -y y } ;.... - (.8) m 6 { M M - } 60 { 8N 7N - } { y - y- } ;.... (.9) 6 m - { M M } { 8N 7N } 60 { y l- y } ; 0... - l (.0) m - { n - 8n n } {y - y. }; 0 M - { N - 8N N } {y -l- y y l }; 0... - (. )... - (.)

- 4 - M {m - m - - m - m } {y - 6y -l - 4y 6y l y } ;... - (.) N { M - 8M M } 0 4 { y - - y y. } ;... K-. (.4) Te relatons (.) (4) can be derved from te results of Albasny and Hosns (97) Fyfe (97) Alberg Nlson and Wals (966) and Saa (970). Te most convenent way for constructng Q s probably troug ts B-splne representaton were as usual. Q ( x ) c ( x x ) l ( t x ) ( t 0 x ) t t < x x. However t s mportant to observe tat te unque exstence of Q can be establsed by sowng tat any of te four ( ) () lnear systems obtaned by usng eter of te relatons (.) (.) (.4) or (.) togeter wt te four equatons derved from te end condtons of Q > s non-sngular. For example f te lnear system correspondng to (.) as a unque soluton M ; 0 l... ten (.4) and (.6) gve te parameters N. ; 0l... and Q can be constructed n any nterval [x - x ] by ntegratng Q (4) ) _ { N _ x) N -x _ ) } four tmes wt respect to x and settng (.) Q ) y and Q () ) M ;.

- - for te determnaton of te four constants of ntegraton. Smlarly f te lnear system correspondng to (.) as a unque soluton m ; 0... ten (.) and (.8) gve te parameters M ; 0... and Q can be constructed n any nterval [x - x ] by use of te two pont quntc Hermte nterpolaton formula. Smlar arguments establs te unque exstence of Q n te two cases were te lnear systems correspondng to (. 4) and (.) are non-sngular. Te followng two lemmas are also needed for te dervaton of te results gven n Secton : Lemma.. If y C 6 [ab] ten for x [x - x ] ;... (r) Q ) (r) y ) l A r - r max 0 m y 6 0( r ); r0..6 (.6) were te A r are constants ndependent of. Lemma.. Let λ m - y. If y C 7 [ a b] ten were λ - 6λ - 66λ 6λ λ β ; - (.7) β 6 y (7) ; -...-. (.8) Lemma. can be establsed from te representaton of Q by means of te two pont quntc Hermte nterpolaton formula by usng a trval generalzaton of a result due to Hall (968: p.4). Lemma. follows easly from (.) by Taylor seres expanson about te pont x. End Condtons We let Q be a quntc splne nterpolatng te values y. y ) ; 0 l... at te equally spaced nots (.) and assume tat 6. As before we use te abbrevatons (.).

- 6 - We consder end condtons of te form () 0.. ; y a 60 m m m m y a 60 m m m m l l γ β α γ β α and see to determne te scalars αβγ and a ; 0... so tat Q exsts unquely and Q (r) _ y (r) 0 ( 6-r ) ; r 0.... (.) For ts we let λ -m -y ; 0... and assume tat y C 7 [ab]. Ten te equatons (.) and (.) gve (.4) β λ αλ βλ γλ β λ αλ βλ 4 γλ... ; β λ λ 6 λ 66 λ 6 λ β 4 γλ βλ αλ λ 0 β γλ βλ αλ 0 λ κ κ κ κ κ κ κ κ κ κ κ ι were ) 4. ( 0 ; y a 60 ) y ( y a 60 ) y ( ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) γ βγ αγ γ β γ βγ αγ γ β 6 y (7) ;... -. (.) ß Te matrx of coeffcents n (.) s te matrx of te ( l) ( I) lnear system wc determnes te parameters m of te quntc splne Q. We denote ts matrx by A let I be te set I {(α.β. γ) ; det A 0 } (.6)

- 7 - and for any (α ß ϒ) I we assume tat tere exsts a number M ndependent of suc tat A - M. Ten for any (α ß ϒ) I Q exsts unquely and snce max o max o λ A β Lemma. sows tat Q (r) - y (r) 0( 6-r ) ; r 0... only f ß 0 ( ); 0 -. Teorem.. Let Q be an nterpolatory quntc splne wc agrees wt y C 7 [ab] at te equally spaced nots (.) and satsfes end condtons of te form (. ) were (α ß ϒ). I. Ten Q (r) - y (r) 0( 6-r ) ; 0 only f n (. ) Proof. a a a a a a 0 7 α β γ 00 6 α 0 β γ 00 0 α 0 β 60 γ 00 60 α 60 β 0 γ 4 7 0 α β 0 γ α β γ By Taylor seres expanson we fnd tat (.7) β 0 ( ); 0 - only f te scalars α β γ and a ; 0... satsfy te relatons (.7). we fnd tat β β More specfcally wen te relatons (.7) old (6) (0 α β γ) y F 6 60 (6) (0 α β γ) y F 60 0 6 (.8)

- 8 - were y (7) F {6880 40( α γ ) 0400 a a 8( a o a 4 ) 87 a }; 0 l -. (.9) Defnton.. A quntc splne Q wc nterpolates te values y ; 0..... at te equally spaced nots (.) and satsfes end condtons of te form (.) wt te a.; 0... gven by (.7) wll be called an E(α ß ϒ) quntc splne. For any (α ß ϒ) I an E(α ß ϒ) quntc splne exsts unquely and by Teorem. f y y ) were y C 7 [ab] ten (r) Q _ y (r) 0( 6-r ) ; r 0 0.... In general for any (α ß ϒ) I ß 0( ); - 0 - and tus m - y 0( ); - 0.... However f te values α ß ϒ are suc tat 0 - α ß ϒ 0 (.0) ten from (.8) ß - 0( 6 ); - 0 -. For ts reason te class of E(α ß ϒ) quntc splnes wose parameters satsfy (.0) s "best" n te sense tat for any member of ts class m - y 0( 6 ); 0.... (.l)

- 9 - Corollary.. Te end condtons of an E(α ß ϒ) quntc splne can be wrtten as m m αm αm βm βm γm γm p ) ap ) βp ) γp p ) ap ) βp ) γp ); ); 0 (.) were p. denotes te quntc polynomal nterpolatng te values y.y l.... y at te ponts x x.... x. Proof. p p p p p p Te proof follows from (.)(7) and te results ) { 7y 00y 00y 00y 7y 4 y } 60 ) { y 6y 0y 60y 0y 4 y } 60 ) {y 0y 0y 60y y 4 y } 60 ) { y y 60y 0y 0y 4 y } 60 4 ) {y 0y 60y 0y 6y 4 y } 60 ) { y 7y 00y 00y 00y 4 7y 60 (.) }. Corollary. sows tat te end condtons of te E(000) quntc splne are m m p ); 0 p );. (.4) Clearly (000) I and terefore an E(000) quntc splne exsts unquely. In fact t s easy to see tat n ts case te (-4) (-4) matrx A of (.) s suc tat A.

- 0 - Te corollary also sows tan an E ( ß ϒ) quntc splne can be nterpreted as an nterpolatory quntc splne wt end condtons m m p p ( ) ( ) 4 ( ( x x ); );. (. ) Smlarly te E (α ß ϒ) and E(α ß ) quntc splnes can e nterpreted as nterpolatory quntc splnes wt end condtons respectvely m p ); m p ); and m p ); 4 m p ); 4. (.6) (.7) Te unque exstence of eac of te E( ß ϒ E( α ϒ) and E(αß ) quntc splnes can be establsed easly by consderng te lnear systems for te m 's derved from te consstency relaton (.) and te end condtons (.)- (.7). However te exstence of te E(α ϒ) ) and E(αβ ) splnes can only be establsed under te assumptons 7 and 9 respectvely. Te corollares stated below establs varous alternatve representatons for te end condtons of an E(α ßϒ) Quntc splne. Tey are establsed by usng te quntc splne denttes lsted n Secton and te expressons for te dervatves p () ); 4 of te nterpolatng polynomal p ). Altoug te algebra nvolved n te dervaton of tese results s very laborous te proofs are oterwse elementary and for ts reason te detals are omtted.

- - Corollary.. Te end condtons of an E(α ß ϒ) quntc splne can be wrtten as A o Δ N 4 A Δ N A Δ N A Δ N 0 A o N A 4 N A N A 0 were N Q (4) ) N 0; (.8) A A A A o ( α β γγ) (69 7 α β γγ 0(7 α 7 β 9 γ9γ 0(0 α β γ) (.9) and A V are respectvely te forward and bacward dfference operators. Corollary. sows tat te fve E(α ß ϒ) quntc splnes lsted below ave partcularly smple representatons of te form Δ n N n N - 0. Splne End condtons E Δ Ν N - 0; 0 E() Δ N N - 0; E(9 9 ) Δ N N - 0; 0 (.0) E(7 9) Δ 4 Ν 4 Ν _ 0; 0 E(6) Δ N N - 0 ; 0 Te unque exstence of eac of tese fve E(α ß ϒ) quntc splnes can be establsed easly by consderng te lnear system

- - for te parameters N derved by usng te end condtons (.0) and te conastency relatons (.). However te exstence of te E() E(79) and E(6) splnes can be establsed only under te assumptons 7 7 and 8 respectvely. We note tat te parameters of eac of te E(99 ) E(79) and E( 66) quntc splnes satsfy (.0) and tus for eac of tese tree splnes max m - y 0 ( n ) (.) o were n 6 rater tan n. Corollary.. Te end condtons of an E(α ß ϒ) quntc splne can B B {m {m p () p )} 0 () )} 0; 0 (.) be wrtten as were M Q () ) B o ( 8α 7β 8γ-) B ( 77 α 8 β 67 γ 8) B ( 8 α 77 β 8 γ 67) B ( 7α 8β γ 8) and p denotes te quntc polynomal nterpolatng te values (.) y y.... y at te Ponts x x l.... x. Wen α 87/888 ß - 890/888 and ϒ 7/888 ten n (.) B B B 0 It follows from Corollary tat te end condtons of te E 87 888 890 888 7 888 can be wrtten as M M p p ( ) ( ) ( x ( x ); ); 0. (.4 )

- - Te unque exstence of ts splne follows at once by consderng te lnear system for te parameters M. derved from te consstency relatons (.) and te end condtons (.4). Corollary.4. Te end condtons of an E(α ß ϒ) quntc splne can be wrtten as C C {n {n p were n Q () ) C o α γ -8 C 8 α β C α 8 β C β 8 γ - () () p 6 γ - 6 6 γ - 6 )} 0 )} 0; 0 and p denotes te quntc polynomal nterpolatng te values (.) (.6) y y... y at te ponts x x... x. Wen α 74/78 ß - 040/78 and ϒ 9/78 ten n (.6) C C C 0. It follows from Corollary.4 tat te end condtons of te 74 040 9 E splne can be wrtten as 78 78 78 n n ( ) p ( ) p ( x ( x ); 0 );. Te unque exstence of ts splne follows at once by consderng te lnear system for te parameters n derved from te consstency relatons (.4) and te end condtons (.7). (.7 )

- 4 - Corollary.. Te end condtons of an E(α ß ϒ) quntc splne can be wrtten as D D {N {N p (4) p )} (4) 0 )} 0; 0 (.8) were NQ (4) ) D o (8 α β γ -9) D ( α 8 β 7 γ - 6) D (8 α β 6 γ -7) D ( α β 9 γ - ) and p denotes te quntc polynomal nterpolatng te values (.9) y y l... y at te ponts x x l.... x. Wen α 9/8 ß - 79/8 and ϒ 7/8 ten n (.9) D D D 0. It follows from Corollary 9 tat te end 9 79 7 condtons of te E can be wrtten as 8 8 8 N N p p (4) (4) ); 0 );. (.0) Te unque exstence of ts splne follows at once by consderng te lnear system for te N derved from te consstency relatons (.) and te end condtons (.0). Let d denote te ump dscontnuty of Q () at te not x. Ten from (.) d Q( )( X ) - Q ( ) ( X -) Δ N -l N ; -. (.)

- - E It follows at once from (. 0) tat te fft dervatves of te and E( ) quntc splnes are contnuous respectvely at te nots x ; - - and x ; J - -. Tese propertes are te specal cases G G G 4 0 and G G G 4 0 of te general result contaned n te followng corollary. Ts result s establsed easly by usng (.) and te result of Corollary.. Corollary.6. Let Q be an E(α ß ϒ) quntc splne and let d denote te ump dscontnuty of Q () at x x Ten G G d d G G d d G G d d G G 4 4 d d 0; 0; (.) were G G G G 4 6 76 09 6 α 9 α 49 α 76 α 6 β β 66 β 49 β 9 γγ γ 49 γ9 66 γ6 (.) A number of propertes of specal nterest wc emerge from te result of Corollary.6 are lsted below. Splne Property satsfed by te d ' s E d d d - d - 0 E( ) d d d - d - 0 E(9 9 ) d d d and d - d - d - E(7 9) Δ d d -. 0; (.4) E( 6 ) Δ d d - 0;

- 6-4. Numercal results In Table.. we present numercal results obtaned by tang y ) exp) x 0. 0 ; 0... 0 (4.) and constructng varous E(α ß ϒ) quntc splnes. Te splnes consdered are te E(000) quntc splne and te fve splnes of (..0). Te results lsted are values of te absolute error exp) - Q) (4.) computed at varous ponts between te nots. For comparson purposes we also lst results computed by constructng te natural quntc splne (N.Q.S.) wt nots (4.) nterpolatng te functon y) exp) at te nots. Te numercal results ndcate te serous damagng effect tat te natural end condtons Q (r) 0 ) - Q (r) 0 ) 0; r 4 (4.) ave upon te qualty of te approxmaton and demonstrate te consderable mprovement n accuracy obtaned by usng end condtons of te type consdered n te present paper nstead of (4.). Te results also sow tat as predcted by te teory te 'best' E(α ß ϒ) quntc splnes correspond to values α ß ϒ tat satsfy (.0).

TABLE Values of exp) - Q) X N.Q.S. E(000) E E( ) E(99) E(79) E(6 ) 0.0 0.9 0-0.7 0-9 0.X0-9 0.x0-8 - 0.70 0-0.84 0 0.7x0-0.0 0. 0-0.78 0-9 0. 0-0.8x0-8 0.84 0-0.x0-0.X0-0.07 0. x 0-0.7X0-9 0.6x0-0 0.4x0-9 0.x0 - - 0.94x0 0.69x0-0.09 0.X0-6 0.X0-9 0.4xl0 0 0.76 l0-0 0 l6-0.xl0-0.6xl0-0. 0.9 0-7 0.9X0-0 0.4x0-0.9X0-0.l0 l0-0; 0-0.X0-0.6 0.X0-8 0.40X0-0.xl0-0. l0-0.6xl0 0.6xl0-0.xl0-0.6 0.6X0-8 0.98X0-0.ll l0-0.x0-0.7 0-0.7X0-0.7X0-0.9 0.X0-0.4 0-8 0.xl0-9 0.49 l0-9 0. l0-0.0 l0-0.4x0-0.96 0.X0-0. x 0-8 0.6x0-9 0.xl0 8 0.7 0 - O.4 l0-0.94xl0-0.98 0.9 0-0. 0-8 0.xl0-9 0.8x0-8 0.x0-0 - 0.8x0 0.4 l0-0.99 0.77X0-0.9 0-9 0.4 l0-9 0.6xl0-8 0.0xl0-0 0.X0-0.l0 l0 -

- 8 - R E F E R E N C E S AHLBERG J.H. NILSON E.N. & WALSH J.L. 967. Te teory of aplnes and ter applcatons. London: Academc Press ALBASINY E.L. & HOSKINS W.D. 97. Te Numercal calculaton of odd degree polynomal splnes wt equ-spaced nots. J.Inst.Mats Apples 7 84-97. BEHFOROOZ G.H. & PAPAMICHAEL N.979. End condtons for cubc splne nterpolaton J. Inst.Mats Apples -66. FYFE D.J. 97. Lnear dependence relatons connectng equal nterval Nt degree splnes and ter dervatves. J.Inst.Mats Apples 7 98-406. HALL C.A. 968. On error bounds for splne nterpolaton. J.Approx.Teory 7 4-47. SAKAI M. 970. Splne nterpolaton and two-pont boundary value problems. Memors of Faculty of Scence Kyasu Unversty Seres A XXIV 7-4. XB 640 X