# INTERPOLATION Interpolation is a process of finding a formula

date post

01-Jan-2017Category

## Documents

view

227download

1

Embed Size (px)

### Transcript of INTERPOLATION Interpolation is a process of finding a formula

INTERPOLATION

Interpolation is a process of finding a formula (oftena polynomial) whose graph will pass through a givenset of points (x, y).

As an example, consider defining

x0 = 0, x1 =

4, x2 =

2and

yi = cosxi, i = 0, 1, 2

This gives us the three points

(0, 1) ,4 ,

1sqrt(2)

,

2 , 0

Now find a quadratic polynomial

p(x) = a0 + a1x+ a2x2

for which

p(xi) = yi, i = 0, 1, 2

The graph of this polynomial is shown on the accom-panying graph. We later give an explicit formula.

Quadratic interpolation of cos(x)

x

y

/4 /2

y = cos(x)y = p2(x)

PURPOSES OF INTERPOLATION

1. Replace a set of data points {(xi, yi)} with a func-tion given analytically.

2. Approximate functions with simpler ones, usually

polynomials or piecewise polynomials.

Purpose #1 has several aspects.

The data may be from a known class of functions.Interpolation is then used to find the member of

this class of functions that agrees with the given

data. For example, data may be generated from

functions of the form

p(x) = a0 + a1ex + a2e

2x + + anenx

Then we need to find the coefficientsnajobased

on the given data values.

We may want to take function values f(x) givenin a table for selected values of x, often equally

spaced, and extend the function to values of x

not in the table.

For example, given numbers from a table of loga-

rithms, estimate the logarithm of a number x not

in the table.

Given a set of data points {(xi, yi)}, find a curvepassing thru these points that is pleasing to the

eye. In fact, this is what is done continually with

computer graphics. How do we connect a set of

points to make a smooth curve? Connecting them

with straight line segments will often give a curve

with many corners, whereas what was intended

was a smooth curve.

Purpose #2 for interpolation is to approximate func-

tions f(x) by simpler functions p(x), perhaps to make

it easier to integrate or differentiate f(x). That will

be the primary reason for studying interpolation in this

course.

As as example of why this is important, consider the

problem of evaluating

I =Z 10

dx

1 + x10

This is very difficult to do analytically. But we will

look at producing polynomial interpolants of the inte-

grand; and polynomials are easily integrated exactly.

We begin by using polynomials as our means of doing

interpolation. Later in the chapter, we consider more

complex piecewise polynomial functions, often called

spline functions.

LINEAR INTERPOLATION

The simplest form of interpolation is probably thestraight line, connecting two points by a straight line.

Let two data points (x0, y0) and (x1, y1) be given.

There is a unique straight line passing through these

points. We can write the formula for a straight lineas

P1(x) = a0 + a1x

In fact, there are other more convenient ways to write

it, and we give several of them below.

P1(x) =x x1x0 x1

y0 +x x0x1 x0

y1

=(x1 x) y0 + (x x0) y1

x1 x0= y0 +

x x0x1 x0

[y1 y0]

= y0 +

y1 y0x1 x0

!(x x0)

Check each of these by evaluating them at x = x0and x1 to see if the respective values are y0 and y1.

Example. Following is a table of values for f(x) =tanx for a few values of x.

x 1 1.1 1.2 1.3tanx 1.5574 1.9648 2.5722 3.6021

Use linear interpolation to estimate tan(1.15). Then

use

x0 = 1.1, x1 = 1.2

with corresponding values for y0 and y1. Then

tanx y0 +x x0x1 x0

[y1 y0]

tanx y0 +x x0x1 x0

[y1 y0]

tan (1.15) 1.9648 + 1.15 1.11.2 1.1 [2.5722 1.9648]

= 2.2685

The true value is tan 1.15 = 2.2345. We will want

to examine formulas for the error in interpolation, to

know when we have sufficient accuracy in our inter-

polant.

x

y

1 1.3

y=tan(x)

x

y

1.1 1.2

y = tan(x)y = p1(x)

QUADRATIC INTERPOLATION

We want to find a polynomial

P2(x) = a0 + a1x+ a2x2

which satisfies

P2(xi) = yi, i = 0, 1, 2

for given data points (x0, y0) , (x1, y1) , (x2, y2). One

formula for such a polynomial follows:

P2(x) = y0L0(x) + y1L1(x) + y2L2(x) ()with

L0(x) =(xx1)(xx2)(x0x1)(x0x2), L1(x) =

(xx0)(xx2)(x1x0)(x1x2)

L2(x) =(xx0)(xx1)(x2x0)(x2x1)

The formula () is called Lagranges form of the in-terpolation polynomial.

LAGRANGE BASIS FUNCTIONS

The functions

L0(x) =(xx1)(xx2)(x0x1)(x0x2), L1(x) =

(xx0)(xx2)(x1x0)(x1x2)

L2(x) =(xx0)(xx1)(x2x0)(x2x1)

are called Lagrange basis functions for quadratic in-

terpolation. They have the properties

Li(xj) =

(1, i = j0, i 6= j

for i, j = 0, 1, 2. Also, they all have degree 2. Their

graphs are on an accompanying page.

As a consequence of each Li(x) being of degree 2, we

have that the interpolant

P2(x) = y0L0(x) + y1L1(x) + y2L2(x)

must have degree 2.

UNIQUENESS

Can there be another polynomial, call it Q(x), forwhich

deg(Q) 2Q(xi) = yi, i = 0, 1, 2

Thus, is the Lagrange formula P2(x) unique?

Introduce

R(x) = P2(x)Q(x)From the properties of P2 and Q, we have deg(R) 2. Moreover,

R(xi) = P2(xi)Q(xi) = yi yi = 0for all three node points x0, x1, and x2. How manypolynomials R(x) are there of degree at most 2 andhaving three distinct zeros? The answer is that onlythe zero polynomial satisfies these properties, and there-fore

R(x) = 0 for all x

Q(x) = P2(x) for all x

SPECIAL CASES

Consider the data points

(x0, 1), (x1, 1), (x2, 1)

What is the polynomial P2(x) in this case?

Answer: We must have the polynomial interpolant is

P2(x) 1meaning that P2(x) is the constant function. Why?First, the constant function satisfies the property ofbeing of degree 2. Next, it clearly interpolates thegiven data. Therefore by the uniqueness of quadraticinterpolation, P2(x) must be the constant function 1.

Consider now the data points

(x0,mx0), (x1,mx1), (x2,mx2)

for some constant m. What is P2(x) in this case? Byan argument similar to that above,

P2(x) = mx for all x

Thus the degree of P2(x) can be less than 2.

HIGHER DEGREE INTERPOLATION

We consider now the case of interpolation by poly-nomials of a general degree n. We want to find apolynomial Pn(x) for which

deg(Pn) nPn(xi) = yi, i = 0, 1, , n ()

with given data points

(x0, y0) , (x1, y1) , , (xn, yn)The solution is given by Lagranges formula

Pn(x) = y0L0(x) + y1L1(x) + + ynLn(x)The Lagrange basis functions are given by

Lk(x) =(x x0) ..(x xk1)(x xk+1).. (x xn)

(xk x0) ..(xk xk1)(xk xk+1).. (xk xn)for k = 0, 1, 2, ..., n. The quadratic case was coveredearlier.

In a manner analogous to the quadratic case, we canshow that the above Pn(x) is the only solution to theproblem ().

In the formula

Lk(x) =(x x0) ..(x xk1)(x xk+1).. (x xn)

(xk x0) ..(xk xk1)(xk xk+1).. (xk xn)we can see that each such function is a polynomial of

degree n. In addition,

Lk(xi) =

(1, k = i0, k 6= i

Using these properties, it follows that the formula

Pn(x) = y0L0(x) + y1L1(x) + + ynLn(x)satisfies the interpolation problem of finding a solution

to

deg(Pn) nPn(xi) = yi, i = 0, 1, , n

EXAMPLE

Recall the table

x 1 1.1 1.2 1.3tanx 1.5574 1.9648 2.5722 3.6021

We now interpolate this table with the nodes

x0 = 1, x1 = 1.1, x2 = 1.2, x3 = 1.3

Without giving the details of the evaluation process,

we have the following results for interpolation with

degrees n = 1, 2, 3.

n 1 2 3Pn(1.15) 2.2685 2.2435 2.2296Error .0340 .0090 .0049

It improves with increasing degree n, but not at a very

rapid rate. In fact, the error becomes worse when n is

increased further. Later we will see that interpolation

of a much higher degree, say n 10, is often poorlybehaved when the node points {xi} are evenly spaced.

A FIRST ORDER DIVIDED DIFFERENCE

For a given function f(x) and two distinct points x0and x1, define

f [x0, x1] =f(x1) f(x0)

x1 x0This is called a first order divided difference of f(x).

By the Mean-value theorem,

f(x1) f(x0) = f 0(c) (x1 x0)for some c between x0 and x1. Thus

f [x0, x1] = f0(c)

and the divided difference in very much like the deriv-

ative, especially if x0 and x1 are quite close together.

In fact,

f 0x1 + x02

f [x0, x1]

is quite an accurate approximation of the derivative

(see 5.4).

SECOND ORDER DIVIDED DIFFERENCES

Given three distinct points x0, x1, and x2, define

f [x0, x1, x2] =f [x1, x2] f [x0, x1]

x2 x0This is called the second order divided difference of

f(x).

By a fairly complicated argument, we can show

f [x0, x1, x2] =1

2f 00(c)

for some c intermediate to x0, x1, and x2. In fact, as

we investigate in 5.4,f 00 (x1) 2f [x0, x1, x2]

in the case the nodes are evenly spaced,

x1 x0 = x2 x1

EXAMPLE

Consider the table

x 1 1.1 1.2 1.3 1.4cosx .54030 .45360 .36236 .26750 .16997

Let x0 = 1, x1 = 1.1, and x2 = 1.2. Then

f [x0, x1] =.45360 .54030

1.1 1 = .86700

f [x1, x2] =.36236 .45360

1.1 1 = .91240

f [x0, x1, x2] =f [x1, x2] f [x0, x1]

x2 x0=

.

*View more*