Chapter 3, Interpolation and

Extrapolation
Interpolation & Extrapolation

(xi,yi)

Find an analytic function f(x)

that passes through given N
points exactly.
 Polynomial Interpolation
• Use polynomial of degree N-1 to fit exactly
with N data points (xi,yi), i =1, 2, …, N.
N 1
f ( x)  c0  c1 x  c2 x 
2
 cN 1 x

• The coefficient ci is determined by a

system of linear equations
 Vandermonde Matrix Equation

1 x1 x2
1 xN 1
1
  c0   y1 
 2 N 1    
1 x2 x2 x2  c1 y
 2
    
 N 1     
1 xN xN  cN 1   yN 
2
x N

• But it is not advisable to solve this system

numerically because of ill-conditioning.
 Condition Number
• cond(A) = ||A|| · ||A-1||

• For singular matrix, cond(A) = ∞

• A linear system is ill-conditioned if cond(A)
is very large.

• Norm || . ||: a function that satisfies (1) f(A)

≥ 0, (2) f(A+B) ≤ f(A) + f(B), (3) f(A) = ||
f(A).
 Norms
• Vector p-norm

 
1/ p
x p
 | x1 |  | x2 | 
p p
| xN | p

• Matrix norm
|| Ax ||
A  sup
x  0 || x ||

sup: supremum
 Commonly Used Norms
• Vector norm
N N
|| x ||1   | x j |, || x ||2   j
x 2

j 1 j 1

|| x ||  max | x j |
j

• Matrix norm
N
|| A ||1  max  | aij |, || A ||2   Where μ is the maximum
j
i 1 eigenvalue of matrix ATA.
N
|| A ||  max  | aij |
i
j 1
 Lagrange’s Formula
• It can be verified that the solution to the
Vandermonde equation is given by the
N
formula below:
N  ( x  x j )
f ( x)   li ( x) yi , li ( x)  N
j  i , j 1

i 1
 ( xi  x j )
j  i , j 1

• li(x) has the property li(xi)=1; li(xk)=0, k≠i.

Joseph-Louis Lagrange (1736-1813)

Italian-French mathematician
associated with many classic
mathematics and physics –
Lagrange multipliers in
minimization of a function,
Lagrange’s interpolation
formula, Lagrange’s theorem in
group theory and number
theory, and the Lagrangian
(L=T-V) in mechanics and
Euler-Lagrange equations.
 Neville’s Algorithm
• Evaluate Lagrange’s interpolation formula
f(x) at x, given the data points (xi,yi).
• Interpolation tableau P
x1: y1 = P1
P12
x2: y2 = P2 P123
P23 P1234
x3: y3 = P3 P234 Pi,i+1,i+2,…,i+n is a polynomial of
P34 degree n in x that passes
through the points (xi,yi),
x4: y4 = P4 (xi+1,yi+1), …, (xi+n,yi+n) exactly.
 Determine P12 from P1 & P2
• Given the value P1 and P2 at x=x1 and x2,
we find linear interpolation
P12(x) = λ(x) P1+[1-λ(x)] P2
Since P12(x1) = P1 and P12(x2) = P2, we
must have
λ(x1) = 1, λ(x2) = 0
so
λ(x) = λ12= (x-x2)/(x1-x2)
 Determine P123 from P12 & P23
• We write
P123(x) = λ(x) P12(x) + [1-λ(x)] P23(x)
P123(x2) = P2 already for any choice of λ(x). We
require that P123(x1)=P12(x1) =P1 and
P123(x3)=P23(x3) =P3, thus λ(x1) = 1, λ(x3) = 0
Or
x  x3
 ( x)  13 
x1  x3
 Recursion Relation for P
• Given two m-point interpolated value P
constructed from point i,i+1,i+2,…,i+m-1,
and i+1,i+2,…,i+m, the next level m+1
point interpolation from i to i+m is a convex
combination:
Pi (i 1)(i  2)...(i  m )   ( x) Pi (i 1)...(i  m 1)  1   ( x)  P(i 1)(i  2)...(i  m )
x  xi  m
where  ( x)  (i )(i  m ) 
xi  xi  m
Use Small Difference C & D
P1 C1,1 = P12-P1 ( xi  x)(Cm,i 1  Dm,i )
Cm1,i 
P12 xi  xi  m1
D1,1 C2,1=P123-P12
P123
P2
D2,1
C1,2 C3,1=P1234-P123
P23
D1,2 P1234
C2,2
P3 P234 D3,1=P1234-P234

C1,3
P34 D2,2=P234-P34

D1,3 ( xi  m1  x)(Cm,i 1  Dm,i )

Dm1,i 
P4 xi  xi  m1
Deriving the Relation among C & D
PA
D1=PA –P0
C2=P-PA

P0
P =λ PA + (1-λ)PB
C1=PB –P0 D2=P-PB
PB

C2  P  PA  P  P0  (PA  P0 ) D2  (C1  D1 )
 PA  (1   )PB  P0  (PA  P0 )
 (1   )(C1  D1 )
 • Evaluate f(3) given 4-points (0,1), (1,2),
(2,3),(4,0).
C0,1=1 Cm1,i  (1  (i )(i  m 1) )(Cm,i 1  Dm,i )
x1=0, y1=1 =P1
C1,1 = 3 x  xi  m1
D0,1=1 (i )(i  m1) 
xi  xi  m1
C2,1= 0
C0,2=2 D1,1= 2

x2=1, y2=2 =P2

C1,2= 2 D2,1= 0 C3,1= - 5/4
D0,2=2
P1234= 11/4
C0,3=3 D1,2= 1 C2,2= -5/3
D3,1= 5/12
x3=2, y3=3 =P3
C1,3= -3/2
D0,3=3 D2,2=5/6

C0,4=0 D1,3 = 3/2 Dm1,i  (i )(i  m 1) (Cm,i 1  Dm,i )

x4=4, y4=0 =P4
D0,4=0
polint( ) Program
polint(), continued
 Piecewise Linear Interpolation
x  x2
P12 ( x)   ( x) y1  1   ( x)  y2 ,  ( x) 
y x1  x2

P23(x)

P12(x) (x2,y2) (x3,y3) P34(x)

(x1,y1) (x4,y4) P45(x)

(x5,y5)

x
Piecewise Polynomial Interpolation

y
Discontinuous derivatives
across segment

P1234(x)
P1234(x)
(x2,y2) (x3,y3) P2345(x)

(x1,y1) (x4,y4) P2345(x)

(x5,y5)

x
 Cubic Spline
• Given N points (xi,yi), i=1,2,…,N, for each
interval between points i to i+1, fit to cubic
polynomials such that Pi(xi)=yi and
Pi(xi+1)=yi+1.
• Make 1st and 2nd derivatives continuous
across intervals, i.e., Pi(n)(xi+1) = P(n)i+1(xi+1),
n = 1 and 2.
• Fix boundary condition to P’’(x1 or N)=0, or
P’(x1 or N ) = const, to completely specify.
 Reading
• NR, Chapter 3

• See also J. Stoer and R. Bulirsch,

“Introduction to Numerical Analysis,”
Chapter 2. Other chapters are also
excellent, but rather mathematical,
references for other topics.
 Problems for Lecture 3 (Interpolation)

1. Use Neville’s algorithm to find the interpolation value at

x = 0, for a cubic polynomial interpolation with 4 points
(x,y) = (-1,1.25), (1,2), (2,3), (4,0). Also write out the
Lagrange’s formula and compute f(0).

2. Given the same 4 points as above, determine the cubic

splines with natural boundary condition (second
derivatives equal to 0 at the boundary). Give the cubic
polynomials in each of the three intervals. Use the
method outlined in NR page 113 to 116, or a direct fit to
cubic polynomials with proper conditions.