8 Hand-Out On Interpolation
8 Hand-Out On Interpolation
Numerical Analysis I
Section B
First Semester A.Y. 2014-2015
Mathematics Division
Institute of Mathermatical Sciences and Physics
University of the Philippines Los Ba nos
September 12, 2014
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 1 / 11
Chapter Two. Interpolation of Functions
Suppose that a mathematical model for the spread of an epidemic
produces that following table showing the number of deaths due to the
disease on t days after the outbreak.
t 0 10 25 50 100
deaths 0 8 4251 20,677 357
What if an expert wants to analyze weekly data?
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 2 / 11
What if we want to get the optimal selling price for a commodity (in terms
of gross sales) and the most ecient way to gather data for a model is to
sell it at varying prices and observing the number of products sold?
Suppose the data gathered is as follows:
Price 70 82.5 137.5 187.50
Number of Units Sold 1,250 750 550 272.50
How will you nd the optimal price?
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 3 / 11
A common mathematical problem
Given a set of points (x
i
, f(x
i
)) for i = 0, 1, 2, ..., n where the nodes x
i
s
are distinct values of the independent variable. Then either
approximate the value of f at some value of x not included in the list;
or
determine a function g that mimics the behavior of the data in some
sense.
These problems give rise to a two dierent areas of study:
Interpolation and Approximation.
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 4 / 11
Interpolation vs Approximation
Interpolation
the function g is determined by requiring zero errors at the nodes, i.e.,
f(x
i
) g(x
i
) = 0, i = 0, 1, ..., n
Approximation
the function g is chosen such that some measure of error is minimized, for
example
n
i=0
(f(x
i
) g(x
i
))
2
< .
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 5 / 11
Consider the function f(x) = e
x
.
Using the nodes 2, 1, 0, 1, the interpolating polynomial is
P(x) =
x
3
+ 3e
x
2
+ x 2
x + e
3
x
2
+ 3x + 2
x
6e
2
+
3e
2
x
3
+ 2x
2
x 2
+ x
6e
2
.
While an approximating polynomial for f is
Q(x) = 1 + x +
x
2
2!
+
x
3
3!
.
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 6 / 11
Interpolation, a special case of Approximation
Why focus on interpolation?
Common forms of interpolation
polynomial
piecewise polynomial
rational
trigonometric
exponential
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 7 / 11
Why polynomials?
they are the simplest functions practical for computers
their derivatives and integrals easy to compute and are still
polynomials
for any required accuracy, there is a polynomial approximating f
Theorem
Weierstrass (First) Approximation Theorem.
Let f C[a, b]. Then > 0 there is a polynomial P(x) such that
|f(x) p(x)| < , x [a, b].
there is one and only one interpolating polynomial for a
set of points (x
i
, f(x
i
)) for i = 0, 1, 2, ..., n .
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 8 / 11
Algebraic Interpolating Polynomial
A polynomial P
n
(x) P
n
(x, f, x
1
, x
2
, ..., x
n+1
) of degree no greater
that n that has the form
P
n
(x) = c
0
+ c
1
x + ... + c
n
x
n
and coincides with the data points (x
i
, f(x
i
)) for i = 1, 2, ..., n, n + 1
is called the algebraic interpolating polynomial of f.
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 9 / 11
Unisolvence Theorem
Theorem
There is a unique algebraic interpolating polynomial
P
n
(x, f, x
1
, x
2
, ..., x
n+1
).
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 10 / 11
Ngayon pa lang...
The accuracy of the interpolating polynomial is
guaranteed
ONLY FOR ABSCISSAS between the
least and greatest nodes.
NJA Egarguin (IMSP, UPLB) MATH 174 September 12, 2014 11 / 11