You are on page 1of 4

1

FINITE DIFFERENCES:

Suppose, we are given a set of values: ( ), ( ), ( ), ( )( ) and the


values of are equally spaced: , . The values, in some interval,
represent some function ( ) of which either we do not know or the function is very
complicated. Now we want to estimate the value of ( ), for some intermediate value of
[ in the range: ].

To proceed, we calculate the differences of the -coordinates in the following ways:

Forward Differences

If , , .. are the values of , then, ( ), ( ),. ( ) are the


successive differences between neighbouring -values.

We denote differences in the following way:

, ,

We can think the symbol as a forward difference operator and , .etc. are called the
first forward differences.

The differences between the first forward differences are called second forward differences
and so on.

The second, third and successive higher order differences can be calculated easily.

( ) ( ) ( ) ,

Similarly,

( ) ( ) ( )

( ) ( )

Similarly, other differences of any order can be calculated in a straight forward way.

Note:
We can think of a shift operator defined by the following way:
for any .

So,
2

( ) , and so on.

In general, .

Let us now think of the difference,

( )

Thus we can think,

Now with this, we can easily calculate the higher order differences.

For example,

( ) ( )

The following table demonstrates how different forward differences are computed:

Table for Forward Differences

Interpolation:

As the explicit nature of the function ( ) is not known, the relation is approximated by a
simple function ( ) so that the original ( ) [not known] and the approximated function
( ) agree at the set of given points ( ). This is called interpolation. If the interpolating
function is a polynomial of some degree then it is called polynomial interpolation. There are
3

other kinds of interpolations like trigonometric interpolation etc., depending on what kind of
function is used for interpolation.

We can calculate the errors in polynomial and other kinds of interpolations. We will discuss this
later.

Newtons Formula for Interpolation:


[This is also called Gregory-Newton Interpolation.]

Newtons interpolation is done through a simple polynomial of degree for a set of ( )


equidistant data points: ( ), ( ), ( ),.( ).

Consider, , .

The polynomial function of n-th degree, which passes through the given data points, can be
written as

( ) ( ) ( )( ) ( )( )( ) (1)

Since the above function will pass through the data points, by putting them we can determine
the coefficients.

So we obtain

( )

( )( )

Similarly,

In general, .

To obtain the expression of the polynomial function for any arbitrary value , we set
so that .

Also, we put the values of , , .in (1), we get


4

( )
( ) ( )( ) ( )( ) ( )
(2)

#Interpolation by Newton's Forward Difference Formula


#Program by: A. Kar Gupta, email: kg.abhi@gmail.com
#----------------------------------------------------

x=[0,1,2,3] #x-coordinates as List


y=[1,0,1,10] #y-coordinates as List
d=[] #Empty List for differences
h=1.0
t=(4-x[0])/h #Conversion of variable for x=4
coeff=t
sum=y[0]
k=1
for i in range(len(y),1,-1):
for j in range(i-1):
diff=y[j+1]-y[j]
d.append(diff) # Appended to difference List
sum=sum+coeff*d[0]
coeff=coeff*(t-k)/(k+1) # Updating the Coeff
k=k+1
y=d
d=[] # Emptying the list for next round
print 'Interpolated Value = ', sum

You might also like