Interpolation with cubic splines

  • Context: MHB 
  • Thread starter Thread starter evinda
  • Start date Start date
  • Tags Tags
    Cubic Interpolation
Click For Summary
SUMMARY

The discussion centers on the proof of a theorem regarding cubic spline interpolation, specifically the existence and uniqueness of a function \( s \in S_{3}(P) \) that interpolates a function \( f \in C^{1}([a,b]) \) at uniform partitions \( P \). Participants analyze the derivation of the cubic spline formula, focusing on the constants \( c_{i1} \) and \( c_{i0} \) within the context of the linear Lagrange polynomial. The correct formulation of these constants is crucial for achieving the desired spline function, and participants confirm the validity of the calculated \( c_{i1} \) while exploring alternative representations for \( c_{i0} \).

PREREQUISITES
  • Understanding of cubic spline interpolation and its mathematical foundations
  • Familiarity with linear Lagrange polynomials and their applications
  • Knowledge of continuity conditions for piecewise functions
  • Basic calculus, including integration and differentiation of functions
NEXT STEPS
  • Study the derivation of cubic spline formulas in detail, focusing on boundary conditions
  • Learn about the implementation of cubic splines in numerical computing libraries such as SciPy
  • Explore the differences between cubic splines and other interpolation methods like polynomial interpolation
  • Investigate the computational complexity and efficiency of cubic spline algorithms
USEFUL FOR

Mathematicians, computer scientists, and engineers involved in numerical analysis, particularly those working with interpolation techniques and spline functions.

evinda
Gold Member
MHB
Messages
3,741
Reaction score
0
Hello! :) I am looking at the proof of this theorem:
Let $f \in C^{1}([a,b]),P:a=x_{0}<x_{1}<...<x_{n}=b $ uniform partition of $[a,b]$.Then there is exactly one function $s \in S_{3}(P)$ so that $s(x_{i})=f(x_{i}),i=0,...,n$ and $s,s',s''$ continuous at $x_{i}$.Also,$s'(x_{0})=f'(x_{0}),s'(x_{n})=f'(x_{n})$.

but at some point I got stuck :confused:

Let $s'_{i}=s'(x_{i}),s''_{i}=s''(x_{i}),s_{i}=s(x_{i}),y_{i}=f(x_{i})$

From the linear Lagrange polynomial,we find that:
$$s''(x)=\frac{1}{h}[s''_{i}(x-x_{i-1})-s''_{i-1}(x-x_{i})] ,x \in [x_{i-1},x_{i}]$$

Then,integrating two times,we get:
$$s(x)=\frac{1}{6h}[s''_{i}(x-x_{i-1})^{3}-s''_{i-1}(x-x_{i})^{3}]+c_{i1}x+c_{i0}$$
Then using the conditions $s_{i-1}=y_{i-1},s_{i}=y_{i}$, we want to find the constants $c_{i1},c_{i0}$ .In my textbook,there is only the solution,which is $s(x)=\frac{1}{6h}[s''_{i}(x-x_{i-1})^{3}-s''_{i-1}(x-x_{i})^{3}]+(\frac{y_{i}}{h}-s''_{i}\frac{h}{6})(x-x_{i-1})-(\frac{y_{i-1}}{h}-s''_{i-1}\frac{h}{6})(x-x_{i})$

I tried to find the constants..For $c_{i1}$ I found $\frac{y_{i}-y_{i-1}}{h}-\frac{h}{6}(s''_{i}-s''_{i-1})$ ,but replacing the conditions and $c_{i1}$, I don't get the right result! Have I calculated wrong the value of $c_{i1}$ ?
From the condition $s_{i-1}=y_{i-1} \text{ and } c_{i1}$ I get $$c_{i0}=y_{i-1}-\frac{h^{2}}{6}s''_{i-1}-\frac{y_{i}-y_{i-1}}{h}x_{i-1}+\frac{h}{6}(s''_{i}-s''_{i-1})x_{i-1}$$
and from the condition $s_{i}=y_{i} \text{ and } c_{i1}$ I get $$c_{i0}=y_{i}-\frac{h^{2}}{6}s''_{i}-\frac{y_{i}-y_{i-1}}{h}x_{i}+\frac{h}{6}(s''_{i}-s''_{i-1})x_{i}$$
 
Physics news on Phys.org
Hey evinda! :D

evinda said:
Hello! :) I am looking at the proof of this theorem:
Let $f \in C^{1}([a,b]),P:a=x_{0}<x_{1}<...<x_{n}=b $ uniform partition of $[a,b]$.Then there is exactly one function $s \in S_{3}(P)$ so that $s(x_{i})=f(x_{i}),i=0,...,n$ and $s,s',s''$ continuous at $x_{i}$.Also,$s'(x_{0})=f'(x_{0}),s'(x_{n})=f'(x_{n})$.

but at some point I got stuck :confused:

Let $s'_{i}=s'(x_{i}),s''_{i}=s''(x_{i}),s_{i}=s(x_{i}),y_{i}=f(x_{i})$

From the linear Lagrange polynomial,we find that:
$$s''(x)=\frac{1}{h}[s''_{i}(x-x_{i-1})-s''_{i-1}(x-x_{i})] ,x \in [x_{i-1},x_{i}]$$

Then,integrating two times,we get:
$$s(x)=\frac{1}{6h}[s''_{i}(x-x_{i-1})^{3}-s''_{i-1}(x-x_{i})^{3}]+c_{i1}x+c_{i0}$$
Then using the conditions $s_{i-1}=y_{i-1},s_{i}=y_{i}$, we want to find the constants $c_{i1},c_{i0}$ .In my textbook,there is only the solution,which is $s(x)=\frac{1}{6h}[s''_{i}(x-x_{i-1})^{3}-s''_{i-1}(x-x_{i})^{3}]+(\frac{y_{i}}{h}-s''_{i}\frac{h}{6})(x-x_{i-1})-(\frac{y_{i-1}}{h}-s''_{i-1}\frac{h}{6})(x-x_{i})$

I tried to find the constants..For $c_{i1}$ I found $\frac{y_{i}-y_{i-1}}{h}-\frac{h}{6}(s''_{i}-s''_{i-1})$ ,but replacing the conditions and $c_{i1}$, I don't get the right result! Have I calculated wrong the value of $c_{i1}$ ?

Nope. Your value of $c_{i1}$ is all good. (Nod)
Note that the given solution has factors $(x-x_i)$ and $(x-x_{i-1})$.
If you bundle the factors of $x$ from the solution, you get your $c_{i1}$.
From the condition $s_{i-1}=y_{i-1} \text{ and } c_{i1}$ I get $$c_{i0}=y_{i-1}-\frac{h^{2}}{6}s''_{i-1}-\frac{y_{i}-y_{i-1}}{h}x_{i-1}+\frac{h}{6}(s''_{i}-s''_{i-1})x_{i-1}$$
and from the condition $s_{i}=y_{i} \text{ and } c_{i1}$ I get $$c_{i0}=y_{i}-\frac{h^{2}}{6}s''_{i}-\frac{y_{i}-y_{i-1}}{h}x_{i}+\frac{h}{6}(s''_{i}-s''_{i-1})x_{i}$$

Well... I guess the solution should be of the form:
$$... + C(x-x_i) + D(x-x_{i-1}) + E$$
instead of
$$... + c_{i1}x+c_{i0}$$
I haven't checked, but presumably your $c_{i0}$ will be equal to $-Cx_i - Dx_{i-1}+E$.
 

Similar threads

  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 18 ·
Replies
18
Views
3K
  • · Replies 1 ·
Replies
1
Views
1K
Replies
2
Views
2K
Replies
2
Views
2K
  • · Replies 7 ·
Replies
7
Views
2K
  • · Replies 9 ·
Replies
9
Views
3K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 5 ·
Replies
5
Views
2K
  • · Replies 13 ·
Replies
13
Views
3K