Interpolation

Author: dr inż. Robert Szmurło

1 / 14
 Interpolation

●
In the mathematical field of numerical
analysis, interpolation is a method of
constructing new data points within the
range of a discrete set of known data points.

●
In engineering and science, one often has
a number of data points, obtained by sampling
or experimentation, which represent
the values of a function for a limited number
of values of the independent variable.
It is often required to interpolate (i.e., estimate) the value of that function for
an intermediate value of the independent variable.

2 / 14
 Example

●
Polynomial interpolation – the order of the polynomial depends on the
number of input points (support points). If n is the number of points then the
order polynomial is n-1.

3 / 14
 Types of interpolation

●
Linear interpolation (the interpolating function depend linearly on the
parameters ai)

– a) polynomial interpolation

b) trigonometric interpolation

– c) spline interpolation (with a

special case of cubic spline)

●
Rational interpolation

●
Exponential interpolation

4 / 14
 Polynomial interpolation (Vandermonde matrix)

●
Let we denote the polynomial P(x)

● For n+1 arbitrary support points (xi,fi), i=0,1,…,n and

xi≠xk for i≠k there exists a unique polynomial P with

●
This allows us to construct a system of n linear
equations with n unknowns

● Which can be solved for a0, a1, …, an

– a0 = 1.000, a1 = 1.8333, a2 = -1.0000, a3 = 0.1667
5 / 14
MATLAB code
function L02_lagrange
xin = [0 1 2 4];
yin = [1 2 2 3];

xdense = linspace( min(xin)-0.5, max(xin) + 0.5, 100)

ydense = zeros(100,1);

for i = 1:length(xdense)
ydense(i) = myvandermonde( xdense(i), xin, yin);
end

plot(xin, yin, 'or', xdense, ydense, 'b');

end

function y = myvandermonde(x, xin, yin)

%A = [ 1 xin(1) xin(1)^2 xin(1)^3
% 1 xin(2) xin(2)^2 xin(2)^3
% 1 xin(3) xin(3)^2 xin(3)^3
% 1 xin(4) xin(4)^2 xin(4)^3 ];
n = length(xin);
A = zeros(n);
for r = 1:n
for c = 1:n
A(r, c) = xin(r)^(c-1);
end
end
b = yin';
a = A \ b;

% y = a(1) + a(2) * x + a(3) * x^2 + a(4) * x^3;

y = 0;
for i = 1:n
y = y + a(i) * x^(i-1);
end
end
6 / 14
 Lagrange interpolation

●
Let we denote the polynomial p(x)

● For n+1 arbitrary support points (xi,fi), i=0,1,…,n and

xi≠xk for i≠k there exists a unique polynomial P with

●
To find this polynomial we will construct p(x)
explicitly with the help of polynomials Li(x) for
which:

●
The following Lagrange polynomials satisfy this
condition

7 / 14
 Example

●
Find Lagrange interpolating polynomial

●
Answer

Try to find other Lagrange

polynomials yourself. Check the
solution:

8 / 14
 Example MATLAB code

x = [0 1 2 4]
y = [1 2 2 3]
n = length(x);

% xd = 0:0.04:4; % dense vector of x coordinates - only for graphical presentation

xd = linspace(min(x)-0.5,max(x)+0.5, 50); % generate 50 equally spaced points between -0.5 and 4.5
yd = []; % declare en empty vector (it will be filled in within the loop below
for k=1:length(xd) % iterate over all elements of vector xd
xval = xd(k);
s = 0;
for i = 1:n
p = 1;
for j = 1:n
if (i ~= j)
p = p * (xval - x(j)) / (x(i) - x(j));
end
end
s = s + y(i) * p;
end
yd(k) = s;

end
yd; % just to print the contents of vector yd to
% the screen
plot(x,y, 'or', xd, yd, 'b') % plot x,y with circles,

9 / 14
 Newton interpolation

●
Earlier presented Lagrange interpolation has the following limitations:

– If the number of data has to be increased we cannot reuse previous calculations.

– The computational effort required is significant for a single interpolation.
– The estimation of error in interpolation is not easy.

●
The Newton’s divided-difference (difference quotient) overcomes these
limitations.
●
The Newton’s polynomial is of the following form:

10 / 14
 Newton interpolation – difference quotient

●
Let we define a set of difference quotients (divided differences)

●
Now let we construct the table with difference quotients of growing order as
the column index grow

11 / 14
 Example

●
Find Newton interpolating polynomial

●
Answer

12 / 14
MATLAB code
function L04_newton function a = newton_coeffs(xp, yp)
xp = [0 1 2 4]; n = length(xp);
yp = [1 2 2 3]; A = zeros(n, n+1);
A(:,1) = xp;
xd = linspace( min(xp), max(xp), 50); A(:,2) = yp;
yd = zeros(length(xd), 1); for c = 3:n+1
for r = c-1:n
A(r,c)=(A(r,c-1)-A(r-1,c-1))/(A(r,1)-A(rc+2,1));
%for i=1:length(xd) end
% x = xd(i); end
% yd(i) = 1 + 1*(x-0) - 1/2*(x-0)*(x-1) + 1/6*(x-0)*(x- a = zeros(n,1);
1)*(x-2);
for i = 1:n
%end
a(i) = A(i, i+1);
a = newton_coeffs(xp, yp);
end
yd = newton_polyval(a, xp, xd);
end

plot(xp, yp, 'o', xd, yd, '.');

end

function y = newton_polyval(a, xp, x)

y = zeros(length(x), 1);
for i = 1:length(x) A =
s = 0;
for j=1:length(a) 0 1.0000 0 0 0
p = 1; 1.0000 2.0000 1.0000 0 0
for k=1:(j-1) 2.0000 2.0000 0 -0.5000 0
p = p * (x(i) - xp(k)); 4.0000 3.0000 0.5000 0.1667 0.1667
end
s = s + a(j) * p;
end
y(i) = s;
end
end

13 / 14
 Thank you

Homework

Solve all presented here examples on your own on a sheet of paper.

14 / 14