Scribd
Upload
41 views14 pages

Lec 3.4 5

1. Hermite interpolation uses polynomials that match both the function values and derivative values at interpolation points. Cubic spline interpolation pieces together cubic polynomials on subintervals to approximate functions, with the polynomials required to have continuous first and second derivatives at interval boundaries. 2. To construct cubic splines, coefficients of cubic polynomials on each subinterval are determined such that the function value, first derivative, and second derivative are continuous where polynomials meet. This results in a system of equations that can be solved to find the coefficients. 3. If the function has continuous fourth derivatives, the error of the cubic spline approximation is bounded above by a formula involving the fourth derivative and the maximum subinterval length.

Uploaded by

AaronJay Jimankge
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
41 views14 pages

Lec 3.4 5

1. Hermite interpolation uses polynomials that match both the function values and derivative values at interpolation points. Cubic spline interpolation pieces together cubic polynomials on subintervals to approximate functions, with the polynomials required to have continuous first and second derivatives at interval boundaries. 2. To construct cubic splines, coefficients of cubic polynomials on each subinterval are determined such that the function value, first derivative, and second derivative are continuous where polynomials meet. This results in a system of equations that can be solved to find the coefficients. 3. If the function has continuous fourth derivatives, the error of the cubic spline approximation is bounded above by a formula involving the fourth derivative and the maximum subinterval length.

Uploaded by

AaronJay Jimankge
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

3.

4 Hermite Interpolation
3.5 Cubic Spline Interpolation

1
Hermite Polynomial

Definition. Suppose 𝑓 ∈ 𝐶 1 [𝑎, 𝑏]. Let 𝑥0 , … , 𝑥𝑛 be


distinct numbers in [𝑎, 𝑏], the Hermite polynomial
𝑃(𝑥) approximating 𝑓 is that:
1. 𝑃 𝑥𝑖 = 𝑓 𝑥𝑖 , for 𝑖 = 0, … , 𝑛
𝑑𝑃 𝑥𝑖 𝑑𝑓 𝑥𝑖
2. 𝑑𝑥
=
𝑑𝑥
, for 𝑖 = 0, … , 𝑛

Remark: 𝑃(𝑥) and 𝑓(𝑥) agree not only function values but
also 1st derivative values at 𝑥𝑖 , 𝑖 = 0, … , 𝑛.

2
Theorem. If 𝑓 ∈ 𝐶 1 𝑎, 𝑏 and 𝑥0 , … , 𝑥𝑛 ∈ 𝑎, 𝑏 distinct
numbers, the Hermite polynomial is:
𝑛 𝑛

𝐻2𝑛+1 𝑥 = 𝑓 𝑥𝑗 𝐻𝑛,𝑗 (𝑥) + 𝑓′ 𝑥𝑗 𝐻𝑛,𝑗 (𝑥)


𝑗=0 𝑗=0
Where
𝐻𝑛,𝑗 𝑥 = [1 − 2(𝑥 − 𝑥𝑗 )𝐿′ 𝑛,𝑗 (𝑥𝑗 )]𝐿2𝑛,𝑗 (𝑥)
𝐻𝑛,𝑗 𝑥 = 𝑥 − 𝑥𝑗 𝐿2𝑛,𝑗 𝑥
Moreover, if 𝑓 ∈ 𝐶 2𝑛+2 𝑎, 𝑏 , then
𝑥 − 𝑥0 2 … 𝑥 − 𝑥𝑛 2 2𝑛+2
𝑓 𝑥 = 𝐻2𝑛+1 𝑥 + 𝑓 (𝜉(𝑥))
2𝑛 + 2 !
for some 𝜉 𝑥 ∈ 𝑎, 𝑏 .
Remark:
1. 𝐻2𝑛+1 𝑥 is a polynomial of degree at most 2𝑛 + 1.
2. 𝐿𝑛,𝑗 (𝑥) is jth Lagrange basis polynomial of degree 𝑛.
𝑥−𝑥0 2 … 𝑥−𝑥𝑛 2 2𝑛+2
3. 2𝑛+2 !
𝑓 (𝜉(𝑥)) is the error term.
3
3rd Degree Hermite Polynomial
• Given distinct 𝑥0 , 𝑥1 and values of 𝑓 and 𝑓′ at these
numbers.
2
𝑥 − 𝑥0 𝑥1 − 𝑥
𝐻3 𝑥 = 1 + 2 𝑓 𝑥0
𝑥1 − 𝑥0 𝑥1 − 𝑥0
2
𝑥1 − 𝑥
+ 𝑥 − 𝑥0 𝑓 ′ 𝑥0
𝑥1 − 𝑥0
2
𝑥1 − 𝑥 𝑥0 − 𝑥
+ 1+2 𝑓 𝑥1
𝑥1 − 𝑥0 𝑥0 − 𝑥1
2
𝑥0 − 𝑥
+ 𝑥 − 𝑥1 𝑓 ′ 𝑥1
𝑥0 − 𝑥1

4
Hermite Polynomial by Divided Differences
Suppose 𝑥0 , … , 𝑥𝑛 and 𝑓, 𝑓′ are given at these numbers.
Define 𝑧0 , … , 𝑧2𝑛+1 by
𝑧2𝑖 = 𝑧2𝑖+1 = 𝑥𝑖 , for 𝑖 = 0, … , 𝑛

Construct divided difference table, but use


𝑓 ′ 𝑥0 , 𝑓 ′ 𝑥1 , . . , 𝑓 ′ 𝑥𝑛
to set the following undefined divided difference:
𝑓 𝑧0 , 𝑧1 , 𝑓 𝑧2 , 𝑧3 , … , 𝑓 𝑧2𝑛 , 𝑧2𝑛+1 .
The Hermite polynomial is
2𝑛+1

𝐻2𝑛+1 𝑥 = 𝑓 𝑧0 + 𝑓 𝑧0 , … , 𝑧𝑘 𝑥 − 𝑧0 … (𝑥 − 𝑧𝑘−1 )
𝑘=1

5
Divided Difference Notation for Hermite
Interpolation
• Divided difference notation:
𝐻3 𝑥
= 𝑓 𝑥0 + 𝑓 ′ 𝑥0 𝑥 − 𝑥0
+ 𝑓 𝑥0 , 𝑥0 , 𝑥1 𝑥 − 𝑥0 2
+ 𝑓[𝑥0 , 𝑥0 , 𝑥1 , 𝑥1 ] 𝑥 − 𝑥0 2 (𝑥 − 𝑥1 )

6
Problems with High Order Polynomial Interpolation

• 21 equal-spaced numbers to interpolate


1
𝑓 𝑥 = 2 . The interpolating polynomial
1+16𝑥
oscillates between interpolation points.

7
Cubic Splines
• Idea: Use piecewise polynomial interpolation, i.e,
divide the interval into smaller sub-intervals, and
construct different low degree polynomial
approximations (with small oscillations) on the
sub-intervals.
• Challenge: If 𝑓′(𝑥𝑖 ) are not known, can we still
generate interpolating polynomial with
continuous derivatives?

8
Definition: Given a function 𝑓 on [𝑎, 𝑏] and nodes 𝑎 = 𝑥0 <
⋯ < 𝑥𝑛 = 𝑏, a cubic spline interpolant 𝑆 for 𝑓 satisfies:
(a) 𝑆(𝑥) is a cubic polynomial 𝑆𝑗 (𝑥) on [𝑥𝑗 , 𝑥𝑗+1 ], ∀𝑗 =
0,1, … , 𝑛 − 1.
(b) 𝑆𝑗 𝑥𝑗 = 𝑓(𝑥𝑗 ) and 𝑆𝑗 𝑥𝑗+1 = 𝑓 𝑥𝑗+1 , ∀𝑗 = 0,1, … , 𝑛 −
1.
(c) 𝑆𝑗 𝑥𝑗+1 = 𝑆𝑗+1 𝑥𝑗+1 , ∀𝑗 = 0,1, … , 𝑛 − 2.
(d) 𝑆 ′𝑗 𝑥𝑗+1 = 𝑆 ′𝑗+1 𝑥𝑗+1 , ∀𝑗 = 0,1, … , 𝑛 − 2.
(e) 𝑆 ′′𝑗 𝑥𝑗+1 = 𝑆 ′′𝑗+1 𝑥𝑗+1 , ∀𝑗 = 0,1, … , 𝑛 − 2.
(f) One of the following boundary conditions:
(i) 𝑆 ′′ 𝑥0 = 𝑆 ′′ 𝑥𝑛 = 0 (called free or natural boundary)
(ii) 𝑆 ′ 𝑥0 = 𝑓′(𝑥0 ) and 𝑆 ′ 𝑥𝑛 = 𝑓′(𝑥𝑛 ) (called clamped
boundary)
9
Things to match at interior point 𝑥𝑗+1 :
• The spline segment 𝑆𝑗 (𝑥) is on 𝑥𝑗 , 𝑥𝑗+1 .
• The spline segment 𝑆𝑗+1 (𝑥) is on 𝑥𝑗+1 , 𝑥𝑗+2 .
• Their function values: 𝑆𝑗 𝑥𝑗+1 = 𝑆𝑗+1 𝑥𝑗+1 =
𝑓 𝑥𝑗+1
• First derivative values: 𝑆 ′𝑗 𝑥𝑗+1 = 𝑆 ′𝑗+1 𝑥𝑗+1
• Second derivative values: 𝑆 ′′𝑗 𝑥𝑗+1 = 𝑆 ′′𝑗+1 𝑥𝑗+1
10
Building Cubic Splines
• Define:
𝑆𝑗 𝑥 = 𝑎𝑗 + 𝑏𝑗 𝑥 − 𝑥𝑗 + 𝑐𝑗 (𝑥 − 𝑥𝑗 )2 +𝑑𝑗 (𝑥 − 𝑥𝑗 )3
and ℎ𝑗 = 𝑥𝑗+1 − 𝑥𝑗 , ∀𝑗 = 0,1, … , 𝑛 − 1.
Solve for coefficients 𝑎𝑗 , 𝑏𝑗 , 𝑐𝑗 , 𝑑𝑗 by:
1. 𝑆𝑗 𝑥𝑗 = 𝑎𝑗 = 𝑓(𝑥𝑗 )
2. 𝑆𝑗+1 𝑥𝑗+1 = 𝑎𝑗+1 = 𝑎𝑗 + 𝑏𝑗 ℎ𝑗 + 𝑐𝑗 (ℎ𝑗 )2 +𝑑𝑗 (ℎ𝑗 )3
3. 𝑆 ′𝑗 𝑥𝑗 = 𝑏𝑗 , also 𝑏𝑗+1 = 𝑏𝑗 + 2𝑐𝑗 ℎ𝑗 + 3𝑑𝑗 (ℎ𝑗 )2
4. 𝑆 ′′𝑗 𝑥𝑗 = 2𝑐𝑗 , also 𝑐𝑗+1 = 𝑐𝑗 + 3𝑑𝑗 ℎ𝑗
5. Natural or clamped boundary conditions

11
Solving the Resulting Equations
∀𝑗 = 1, … , 𝑛 − 1
ℎ𝑗−1 𝑐𝑗−1 + 2 ℎ𝑗−1 + ℎ𝑗 𝑐𝑗 + ℎ𝑗 𝑐𝑗+1
3 3
= 𝑎𝑗+1 − 𝑎𝑗 − 𝑎𝑗 − 𝑎𝑗−1
ℎ𝑗 ℎ𝑗−1
Remark: (n-1) equations for (n+1) unknowns {𝑐𝑗 }𝑛𝑗=0 .
Once compute 𝑐𝑗 , we then compute:
𝑎𝑗+1 −𝑎𝑗 ℎ𝑗 2𝑐𝑗 +𝑐𝑗+1
𝑏𝑗 = −
ℎ𝑗 3
𝑐𝑗+1 − 𝑐𝑗
𝑑𝑗 =
3ℎ𝑗
12
Completing the System
• Natural boundary condition:
1. 0 = 𝑆 ′′ 0 𝑥0 = 2𝑐0 → 𝑐0 = 0
2. 0 = 𝑆 ′′ 𝑛 𝑥𝑛 = 2𝑐𝑛 → 𝑐𝑛 = 0
• Clamped boundary condition:
a) 𝑆 ′ 0 𝑥0 = 𝑏0 = 𝑓′(𝑥0 )
b) 𝑆 ′ 𝑛−1 𝑥𝑛 = 𝑏𝑛 = 𝑏𝑛−1 + ℎ𝑛−1 (𝑐𝑛−1 + 𝑐𝑛 ) = 𝑓′(𝑥𝑛 )
Remark: a) and b) gives additional equations:
3
2ℎ0 𝑐0 + ℎ0 𝑐1 = (𝑎1 − 𝑎0 ) − 3𝑓′(𝑥0 )
ℎ0
3
ℎ𝑛−1 𝑐𝑛−1 + 2ℎ𝑛−1 𝑐𝑛 = − 𝑎𝑛 − 𝑎𝑛−1 + 3𝑓′(𝑥𝑛 )
ℎ0

13
Error Bound
If 𝑓 ∈ 𝐶 4 [𝑎, 𝑏], let 𝑀 = max |𝑓 4 (𝑥)|. If 𝑆 is the
𝑎≤𝑥≤𝑏
unique clamped cubic spline interpolant to 𝑓 with
respect to the nodes: 𝑎 = 𝑥0 < ⋯ < 𝑥𝑛 = 𝑏, then
with
ℎ = max 𝑥𝑗+1 − 𝑥𝑗
0≤𝑗≤𝑛−1
5𝑀ℎ4
max |𝑓 𝑥 − 𝑆(𝑥)| ≤ .
𝑎≤𝑥≤𝑏 384

14

You might also like