MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Interpolation

Given the set of tabular values (x 0, y0), (x1, y1), (x2, y2),... ... ... (xn, yn), satisfying the
relation y = f(x) where the explicit nature of f(x) is not known, it is requested to find a
simpler function say φ(x), such that f(x) and φ(x) agree at the set of tabulated points.
Such a process is called interpolation.
 If φ(x) is a polynomial, then the process is called polynomial interpolation and φ(x) is
called the interpolating polynomial.
 Similarly, different types of interpolation arise depending on whether φ(x) is a finite
trigonometric series, series of Bessel functions, etc.
 Application of interpolation:
1. Analyzing experimental data
2. Plotting smooth curves to represent experimental data
3. Derive many different formulas
 The justification for replacing a given function by a polynomial or by a trigonometric
series relies on two theorems proved by Weierstrass (1885).
1. If f(x) is continuous in an interval (a, b) (i.e. a ≤ x ≤ n), then given any ε > 0,
there exits a polynomial p(x) such that
| f(x) – p(x) | < ε for all x in (a, b)
2. Every continuous function of period 2л can be represented by a finite
trigonometric series of the form g(x) = a0 + a1 sin x + a2 sin 2x + ... ... ... … +
an sin nx + b1 cos x + b2 cos 2x +... ... ... bn cos nx
where, | f(x) – g(x) | < δ for all values of x in the interval considered and δ
represents any pre assigned positive quantity.
This means that it is possible to find a polynomial p(x) whose graph remain within
the regain bounded by y = f(x) + ε and y = f(x) - ε for all x between a and b, however
small ε may be.

y = f(x)+ ε

y=f(x)
y = f(x) - ε y=p(x)

x=a x=b

Page 1 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Finites Differences
Assume that we have a table of values (x i, yi), i = 0, 1, 2, …, n of any function y = f(x),
the values of x being equally spaced, i.e. xi = x0 + ih, i = 0,1,2, ...,n. Suppose that we are
required to recover the values of f(x) for some intermediate values of x, or to obtain the
derivative of f(x) for some x in the range x 0 ≤ x ≤ xn. The methods for solution of these
problems are based on the concept of the differences of a function which is discussed
below.
Forward Differences
If y0, y1, y2, ... ..., yn denote a set of values of y, then y 1 - y0, y2 - y1, ... ... ... , yn - yn-1 are
called the differences of y. Denoting these differences by Δy 0, Δy1, Δy2,... ... ... , Δyn-1
respectively, we have,
Δy0 = y1 – y0, Δy1 = y2 – y1, .... ...., Δyn-1 = yn – yn-1
General form: Δyi = yi+1 – yi, where i = 0, 1, 2, 3, ….., n-1
where Δ is called the forward difference operator and Δy0, Δy1, Δy2, ..., Δyn-1 are called
first forward differences.
The difference of the first forward differences are called second forward differences and
denoted by Δ2y0, Δ2y1, Δ2y2, ... .… Similarly, one can define third forward differences,
fourth forward differences, etc.
Δ2y0 = Δy1 – Δy0 = (y2 – y1) – (y1 – y0) = y2 - 2y1 + y0
Δ3y0 = Δ2y1 – Δ2y0 = (y3 – 2y2 + y1) - (y2 - 2y1 + y0) = y3 - 3y2 + 3y1 - y0
 Any higher order difference can easily be expressed in terms of the ordinates, since
the coefficients occurring on the right side are the binomial coefficients.
 Δr yi = Δr-1 yi+1 - Δr-1 yi

n
Δny0 = ∑ (-1)k ck yn-k
Δ y0 = yn – nc1 k=0
n
yn-1 + nc2 yn-2 - nc3 yn-3 ... ...+ (-1)n yo
Forward Difference Table
The following table shows how the forward differences of all orders can be formed:

x y Δ Δ2 Δ3 Δ4
x0 y0
Δy0
x1 y1
Δ2y0
Δy1 Δ3y0
x2 y2
Δ2y1 Δ4y0
Δy2 Δ3y1
x3 y3
Δ2y2
Δy3
x4 y4

The above table is called a diagonal difference table. The first term in the table is y0. It is
called the leading term. The difference Δy0, Δ2y0, Δ3y0... are called the leading
differences.
 There is another type of difference table called horizontal difference table.
Example:

Page 2 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

 A convenient check may be obtained by noting the sum of the entries in any column

x y Δ Δ2 Δ3 Δ4
0 7
2
1 9 32
34 60
2 43 92 24
126 84
3 169 176
302
4 471
equals the differences between the first and last entries in preceding column.
Properties of the operator Δ
1. If c is a constant then Δc = 0.
2. Δ is distributive i.e. Δ [f(x) + g(x)] = Δf(x) + Δg(x)
3. If c is a constant then Δ[c f(x)] = c Δf(x)
4. If m and n are positive integers then Δm Δn f(x) = Δm+n f(x)
5. Δ [f1(x) + f2(x) + …. + fn(x)] = Δ f1(x) + Δ f2(x) + …. + Δ fn(x)
6. Δ [f(x) g(x)] = f(x) Δ g(x) + g(x) Δ f(x)

Theorem: The nth differences of a polynomial of the n th degree are constant when the
value of independent variables are at equal intervals.
Proof: Page-51 [G. Shanker Rao]
 The converse of the above theorem is true i.e. if the nth differences of a tabulated
function are constant and the values of the independent variable are equally spaced
then the function is a polynomial of degree n.
Example 4: Page 52 [G. Shanker Rao]

Error Propagation in a Difference Table

Let y0, y1, y2, …., yn be the values of the function y = f(x) and the value y 5 be effected
with an error ε, such that the erroneous value of y 5 is y5 + ε. In this case the error ε
effects the successive differences and spreads out fanwise as higher orders are formed in
the table. The table given below shown us the effect of the error. [p-69, Rao]
Example 25: Page-69 [G. Shanker Rao]

Backward differences
The differences y1 – y0, y2 – y1 ...., yn – yn-1 are called first backward differences if they
are denoted by y1, y2, ... ... , yn respectively, so that y1 = y1–y0, y2 = y2 – y1, ... ...,
yn = yn – yn-1, where  is called the backward difference operator.
The difference of the first backward differences are called second backward differences
and denoted by 2y0, 2y1, 2y2, ... .… Similarly, one can define third backward
differences, fourth backward differences, etc.

Page 3 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

2y2 = y2 – y1 = (y2 – y1) – (y1 – y0) = y2 - 2y1 + y0

3y3 = 2y3 – 2y2 = (y3 – 2y2 + y1) - (y2 - 2y1 + y0) = y3 - 3y2 + 3y1 - y0 etc.
 General form: yi = yi – yi-1, where i = 1, 2, 3, ….., n-1
 k yi = k-1 yi - k-1 yi-1

Backward Difference Table

The following table shows how the backward differences of all orders can be formed.
x y  2 3 4
x0 y0
x1 y1 y1
x2 y2 y2 2y2
x3 y3 y3 2y3 3y3
x4 y4 y4 2y4 3y4 4y4
The above table is called a horizontal difference table.
Central Differences
The central difference operator δ is defined by the relations -
δy1/2 = y1 – y0, δy3/2 = y2 – y1 .... .... δyn-1/2 = yn – yn-1
Similarly,
δ2y1 = δy3/2 – δy1/2 x y δ δ2 δ3 δ4
δ3y3/2 = δ2y2 – δ2y1
In general, x0 y0
δyi = yi+1/2 – yi- 1/2 δy1/2
δnyi = δn-1yi+1 – δn-1yi- ½ x1 y1 δ2y1
 It is clear from the three tables that in a definite δy3/2 δ3y3/2
numerical case, the same number occurs in the same x2 y2 δ2y2 δ4y2
position whether we use forward, backward or central δy5/2 δ3y5/2
differences. x3 y3 δ y3
2

δy7/2
x4 y4
The Averaging Operator (μ)
The averaging operator μ is defined as
μyi = ½(yi+1/2 + yi - 1/2)
Shift Operator (E)
The shift operator (also called displacement operator) E is defined by the equation
Eyi = yi+1 or E f(x) = f(x+h)
which shows that the effect of E is to shift the functional value yi to the next higher value yi+1.
E2yi = E(Eyi)
Eyi+1 = yi+2
So,
Enyi = yi+n
Eny0 = yn

Page 4 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

En f(x) = f(x+nh)
E-n f(x) = f(x-nh)

D Operator
The D operator is defined as
Dy(x) = d/dx y(x)
Df(x) = d/dx f(x) = f′(x)
D2f(x) = d2/dx2 f(x) = f″(x)

Some Important Relations between different operators

1. E  1 + 
y0 = y1 – y0 = Ey0 - y0 =(E-1)y0
E–1
2. E  E 
Eyn = E(yn+1 – yn ) = Eyn+1 – Eyn = yn+2 – yn+1 = yn+1 = Eyn
E  E
3.   1 – E-1
yn = yn – yn-1 = yn – E-1yn = (1 – E-1)yn
  1 – E-1
4.  -   
 -   (E - 1) – (1-E-1)  E – 2 + E-1
  (E - 1)(1-E-1)  E – 2 + E-1
  -    
5.   E  E  δ E1/2
  (1-E-1)
 E  E(1-E-1)  E – 1  
 E  (1-E-1)E  E – 1  
 δ  E1/2 – E-1/2
δE1/2  E1/2 (E1/2 – E-1/2)  E – 1  
   E  E  δ E1/2
6. δ  E1/2 – E-1/2
δyr = yr+1/2 – yr-1/2
= E1/2yr – E-1/2yr = ( E1/2 – E-1/2 )yr
δ  E1/2 – E-1/2
7.   ½( E1/2 + E-1/2)
yr = ½ ( yr+1/2 + yr-1/2 )
= ½( E1/2 yr + E-1/2 yr) = ½( E1/2 + E-1/2 )yr
  ½(E1/2 + E-1/2)
8. 2 =1 + 1/4 δ2
  ½(E1/2 + E-1/2)
2  ½(E1/2 + E-1/2)2  ¼{ (E1/2 - E-1/2)2 + 4}  ¼( δ 2 + 4)  ¼δ 2 + 1
2 =1 + 1/4 δ2
Home Task:

Page 5 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

9.  -   δ2
10. E1/2   + ½δ
11. (1+ )(1-)  1
12.   E-1
13. δ  E-1/2 

Newton’s Binomial Expansion formula

Page 6 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Interpolation with Equal Intervals

Newton’s Forward interpolation formula

Let y = f(x) be a function which takes the values y 0, y1, y2, …., yn corresponding to the
(n+1) values x0, x1, x2, …, xn of the independent variables. Let the values of x be equally
spaced i.e. xi = x0 + ih where i = 0,1,2, ... ... n and h is the interval of differencing. Let
φ(x) be a polynomial of the nth degree such that y and φ(x) agree at the tabulated points
i.e. f(xi) = φ(xi), i = 0,1,2... ... ... n and at all other points f(x) = φ(x) + R(x) where R(x) is
called error term of the interpolation formula. Ignoring the error term let us assume
f(x) ≈ φ(x) = a0 + a1 (x-x0) + a2 (x-x0) (x-x1) + ... ... + an (x-x0) (x-x1) ... ... (x-xn-1)
The constants a0, a1, ... ..., an can be determined as follows:
putting x = x0 we get
f(x0) ≈ φ(x0) = a0
∴ a0 = y0
putting x = x1 we get
f(x1) ≈ φ(x1) = a0 + a1 (x1 - x0) = y0 + a1h
∴ a1 = (y1 -y0) / h = ∆y0 / h
putting x = x2 we get
f(x2) ≈ φ(x2) = a0 + a1 (x2 - x0) + a1 (x2 - x0) (x2 – x1)
y2 = y0 + ∆y0/h(2h) + a2 (2h)(h)
y2 = y0 + 2(y1 - y0) + a2 (2h2)
y2 - 2y1 + y0
a2 = ----------------------
2h2
2
Δ y0
∴ a2 = -----------------
2! h2
Similarly by putting x = x3, x = x4, ... ... x = xn we get,
Δ3y0 Δ4y0 Δny0
a3 = -----------------, a4 = -----------------, ……., an = -----------------
3! h3 4! h4 n! hn
putting the values of a0, a1, ... ..., an we get
f(x) ≈ φ(x) = y0 + (∆y0/h)(x-x0) + (Δ2y0 / (2! h2)) (x-x0)(x-x1) + ... ...
... ... + (Δny0/ (n! hn)) (x-x0)(x-x1) ... ... (x-xn-1)
writing u = ((x-x0) / h), we get
x-x0 = uh
x-x1 = (x-x0) + (x0-x1) = (x-x0) - (x1-x0) = uh – h = (u – 1)h
Similarly, x-x2 = (u-2)h
x-x3 = (u-3)h
.... ... ... ... ...
x-xn-1 = (u – n +1)h
f(x) ≈ φ(x) = y0 + u(∆y0/1!) + u(u-1)(Δ2y0)/ 2! + ... ...
... ... + u(u-1)(u-2) ... ... (u – n +1) (Δny0)/n!
The above formula is called Newton’s forward interpolation formula.

Page 7 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

 Newton’s forward interpolation formula is used to interpolate the values of y near

the beginning of a set of tabular values.
Example 5: Page-85 [G. Shanker Rao]
Newton-Gregory Backward Interpolation formula
Newton forward interpolation formula cannot be used for interpolating a value of y near
the end of a table of values. For this purpose, we use another formula known as Newton-
Gregory interpolation formula.
Let y = f(x) be a function which takes the values y 0, y1, y2, …., yn corresponding to the
(n+1) values x0, x1, x2, …, xn of the independent variables. Let the values of x be equally
spaced i.e. xi = x0 + ih where i = 0,1,2, ... ... n and h is the interval of differencing. Let
φ(x) be a polynomial of the nth degree such that y and φ(x) agree at the tabulated points
i.e. f(xi) = φ(xi), i = 0,1,2... ... ... n and at all other points f(x) = φ(x) + R(x) where R(x) is
called error term of the interpolation formula. Ignoring the error term let us assume
f(x) ≈ φ(x) = a0 + a1 (x-xn) + a2 (x- xn) (x- xn-1) + ... ... + an( x- xn)(x- xn-1) ... ... (x-x1)
putting x = xn we get
f(xn) ≈ φ(xn) = a0
∴ a0 = yn
putting x = xn-1 we get
f(xn-1) ≈ φ(xn-1) = a0 + a1(xn-1 - xn) = yn + a1(-h)
∴ a1 = (yn –yn-1) / h = yn / h
putting x = xn-2 we get
f(xn-2) ≈ φ(xn-2) = a0 + a1(xn-2 - xn) + a2 (xn-2 - xn)( xn-2 – xn-1)
yn-2 = yn + yn/h (-2h) + a2 (-2h)(-h)
yn-2 = yn – 2 yn + 2 yn-1 + a2 (2h2)
yn - 2yn-1 + yn-2
a2 = ----------------------
2h2
 yn
2

∴ a2 = -----------------
2! h2
Similarly,
3 yn 4yn nyn
a3 = -----------------, a4 = -----------------, …., an = -----------------
3! h3 4! h4 n! hn
putting the values of a0,a1, ... ...an we get
f(x) ≈ φ(x) = yn + (yn/h)(x-xn) + (2yn / (2! h2)) (x-xn)(x-xn-1) + ... ...
... ... + (nyn/ (n! hn)) (x-xn)(x-xn-1) ... ... (x-x1)
writing u = ((x-xn)/h), we get
x-xn = uh
x-xn-1 = (x-xn) + (xn- xn-1) = (u + 1)h
Similarly, x-xn-2 = (u+2)h
x-xn-3 = (u+3)h
.... ... ... ... ...

x-x1 = (u + n -1)h
f(x) ≈ φ(x) = yn + u (yn /1!) + u (u+1) ( 2yn) / 2! + ... ...

Page 8 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

... ... + u (u+1) (u+2) ... ... (u + n -1) (nyn) / n!

The above formula is called Newton’s backward interpolation formula.
Example 10: Page-90 [G. Shanker Rao]
Example: Find the cubic polynomial which takes the following values
y(0) = 1, y(1) = 0, y(2) = 1, y(3) = 10
Solution:
Here, h =1, u = (x-x0) / h = (x - 0) / 1 = x x y ∆ ∆2 ∆3
φ(x) = 1 + x(-1) + (x(x-1)/2!)(2) + (x(x-1)(x-2)/3!)(6) 0 1
= 1 + (-x) + x(x-1) + x(x-1)(x-2) -1
= 1 – x + x2 – x + x3 – x2 – 2x2 + 2x 1 0 2
= x3 – 2x2 + 1 1 6
Which is the polynomial from which we obtained from the 2 1 8
above tabular values. 9
3 10
 Compute f(4)
Solution: h = 1, u = (4-0) / 1 = 4
φ(4) = 1 + 4(-1) + 4( 4 – 1 ) + 4(4 – 1)(4 – 2) = 1 – 4 + 12 + 24 = 33
Again putting 4 in the equation we get,
43 – 2.42 + 1 = 33

 This process of finding the value of y for some value of x outside the given range is
called extrapolation and this example demonstrates the fact that if a tabulated function
is a polynomial then interpolation and extrapolation would give exact values.
 If a tabulated function is other then a polynomial, then extrapolation very far from the
table limits would be dangerous - although interpolation can be carried out very
accurately. Example 3.8: Page-63 [S. S. Sastry]
Example 3.5: Page-63 [G. Shanker Rao]

Page 9 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Central Difference Interpolation Formula

Newton’s forward and backward interpolation formulae are suited for interpolation near
the beginning and end values of the tabulated data respectively. For interpolation near the
middle of a difference table, central difference formulae are preferable.
Gauss Forward Interpolation formula:
xy ΔyΔ2yΔ3yΔ4yΔ5yΔ6yx-3x0-3hy-3Δy-3x-2x0-2hy-2Δ2y-3Δy-2Δ3y-3x-1x0-hy-
1Δ y-2Δ y-3Δy-1Δ y-2Δ y-3x0x0y0Δ y-1Δ y-2Δ y-3Δy0Δ y-1Δ y-
2 4 3 5 2 4 6 3 5

2x+1x0+hy+1Δ y0Δ y-1Δy+1Δ y0x+2x0+2hy+2Δ y+1Δy+2x+3x0+3hy+3

2 4 3 2

The Newton’s forward interpolation formula is

f(x) = y0 + u∆y0 + u(u-1)(Δ2y0)/ 2! + ... ... + u(u-1)(u-2) ... ...(u – n +1) (Δny0)/n!
where, u = ((x-x0)/h) and x > x0.
From the central difference table we have
Δ3y-1 = Δ2y0 – Δ2y-1
Δ2y0 = Δ2y-1 + Δ3y-1
Similarly, Δ3y0 = Δ3y-1 + Δ4y-1
Δ4y0 = Δ4y-1 + Δ5y-1
Substituting the values, we have
f(x) = y0 + u∆y0 + u(u-1)( Δ2y-1+Δ3y-1)/ 2! + u(u-1)(u-2) (Δ3y-1 + Δ4y-1)/3! + …..
= y0 + u∆y0 + u(u-1)( Δ2y-1)/ 2! + (u+1)u(u-1) (Δ3y-1)/3!
+ (u+1)u(u-1)(u-2) (Δ4y-2)/4! + …..
This is called Gauss forward interpolation formula.
Gauss Backward Interpolation formula
We have, Δ2y-1 = Δy0 – Δy-1
Δy0 = Δy-1 + Δ2y-1
Similarly, Δ2y0 = Δ2y-1+ Δ3y-1
Δ3y0 = Δ3y-1 + Δ4y-1
Again, Δ3y-1 = Δ3y-2 + Δ4y-2
Δ4y-1 = Δ4y-2 + Δ5y-2
Now,
y = f(x) = y0 + u(Δy-1-Δ2y-1) + u(u-1)( Δ2y-1-Δ3y-1)/ 2! + u(u-1)(u-2) (Δ3y-1-Δ4y-1)/3! + ….
= y0 + u∆y-1 + u(u+1)( Δ2y-1)/ 2! + (u+1)u(u-1) (Δ3y-2)/3!
+ (u+2)(u+1)u(u-1) (Δ4y-2)/4! + …….
This is called Gauss backward interpolation formula.

Page 10 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

xy ΔyΔ2yΔ3yΔ4yΔ5yΔ6yx-3x0-3hy-3Δy-3x-2x0-2hy-2Δ2y-3Δy-2Δ3y-3x-1x0-hy-1Δ2y-2Δ4y-3Δy-
1Δ y-2Δ y-3x0x0y0Δ y-1Δ y-2Δ y-3Δy0Δ y-1Δ y-2x+1x0+hy+1Δ y0Δ y-
3 5 2 4 6 3 5 2 4

1Δy+1Δ y0x+2x0+2hy+2Δ y+1Δy+2x+3x0+3hy+3

3 2

Stirlings’s Formula
Taking the mean of the two Gauss’s formula (forward and backward), we obtain
u(Δy0+Δy-1) u2Δ2y-1 u(u2-1) Δ3y-1 + Δ3y-2 u2(u2-1) Δ4y-2
y = y0 + ----------------- + ------------ + ---------- . ----------------- + --------------------- + ... ... ...
2 2 3! 2 4!

This formula is called Stirlings’s formula.

Bessel’s Formula
Changing the origin in the Gauss’s backward interpolation formula from 0 to 1, we have,

u(u-1) u(u-1)(u-2)
y = y1 + (u-1)Δy0 + ------------ Δ2y0 + -------------- Δ3y-1+ ... ... ...
2! 3!

Taking the mean of the above formula and the Gauss’s forward interpolation formula, we obtain,

u(u-1) (u- ½)u(u-1)

y = ½ [y0 + y1] + (u- ½)Δy0 + ------- . ½[Δ2y-1+ Δ2y0] + -----------------Δ3y-1+ ... ... ...
2! 3!

This is called Bessel’s formula.

 Gauss’s forward formula is used to interpolate the values of the function for the value
of u such that 0 < u < 1, and Gauss’s backward formula is used to for a negative
values of u which lies between –1 and 0 i.e. –1 < u < 0. Stirling’s formula gives the
most accurate result for –0.25 ≤ u ≤ 0.25. Bessel’s formula is most efficient for ¼ ≤ u
≤¾.

Example 1, 2, 6, 8: Page 123 – 129 [G. Shanker Rao].

Page 11 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Interpolation with Unequal Intervals

The Newton’s forward and backward interpolation formula and central difference
formulae are applicable only when values of n are given at equal intervals. When the
values of independent variable x are given at unequal intervals then we can use the
following formula.
Lagrange’s Interpolation Formula
Let y = f(x) be a function which assumes the values f(x0), f(x1), f(x2), …., f(xn)
corresponding to the values x = x0, x1, x2, …, xn where the values of x are not equispaced.
Since (n+1) values of the function are given corresponding to the (n+1) values of the
independent variable x, we can represent the function y = f(x) by a polynomial in x of
degree n.
Let the Polynomial be
f(x) = a0(x-x1)(x-x2)...(x-xn) + a1(x-x0)(x-x2)....(x-xn) + a2(x-x0)(x-x1)(x-x3)...(x-xn)
+ ... ... + an(x-x0)(x-x1)... ...(x-xn-1) . . . . . . . . . . . . (1)
Each term in equation (1) being a product of n factors in x of degree n.
Putting x = x0, we get,
f(x0) = a0(x0-x1) (x0-x2) ... ... (x0-xn)
f(x0)
∴ a0 = ------------------------------------
(x0-x1) (x0-x2) ... ... (x0-xn)
Putting x = x1, we get,
f(x1) = a1(x1-x0) (x1-x2) ... ... (x1-xn)
f(x1)
a1 = ------------------------------------
(x1-x0) (x1-x2) ... ... (x1-xn)
Similarly Putting x = x2, x = x3, ... ... , x = xn we get
f(x2)
a2 = ------------------------------------
(x2-x0) (x2-x1) ... ... (x2-xn)
... ... ... ... ... ... ... ... ... ... ... ... ... ...
f(xn)
an = ---------------------------------------
(xn-x0) (xn-x1) ... ... (xn-xn-1)
Substituting the values of a0, a1, ... ...an we get
(x-x1) (x-x2) ... ... (x-xn) (x-x0) (x-x2) ... ... (x-xn)
y = f(x) = -------------------------------- f(x0) + ------------------------------------ f(x1)
(x0-x1) (x0-x2) ... ... (x0-xn) (x1-x0) (x1-x2) ... ... (x1-xn)

(x-x0) (x-x1) ... ... (x-xn-1)

+ ... ... + ---------------------------------- f(xn)
(xn-x0) (xn-x1) ... ... (xn-xn-1)
This is called Lagrange’s interpolation formula. This can also be used for equal intervals.
Example 5, 6: Page – 107 [G. Shankar Rao].

Page 12 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Divided Difference
The first divided difference of f(x) between x0 and x1 is denoted by f[x0,x1] and is defined as
f(x0) - f(x1)
f[x0,x1] = ---------------
x0 - x1
and similarly,
f(x1) - f(x2)
f[x1,x2] = --------------- etc.
x1 - x2
The second order divided difference between x0, x1 and x2
f[x0,x1] - f[x1,x2]
f[x0,x1,x2] = --------------------
x0 - x2
th
The (k+1) divided difference
f[x0,x1... ... xk] - f[x1,x2... ... xk+1]
f[x0,x1,... ...xk+1] = ------------------------------------------
x0 – xk+1
Divided Difference Table: Page-98 [G. Shanker Rao].
Example 2: Page-99 [G. Shanker Rao].
Properties of Divided differences
1. Divided differences are symmetric function of their arguments.
f(x0) - f(x1) f(x1) - f(x0)
f[x0,x1] = --------------- = -------------------- = f[x1,x0]
x0-x1 x1-x0
2. The divided difference of the product of a constant and a function is equal to the
product of the constant and the divided difference of the function i.e.
k f[x0,x1] = f k[x1,x0] where k is a constant.
3. The divided difference of the sum (or difference) of two functions is equal to the sum
(or difference) of the corresponding separate divided differences.
If f(x) = g(x) + h(x) then, f[x0,x1] = g[x0,x1] + h[x0,x1]
4. The nth order divided differences of a polynomial of nth degree are constants.
5. The (n+1)th order divided differences of a polynomial of nth degree will be zero.
Newton’s General Divided Difference Formula
Let y = f(x) be a function which assumes the values f(x0), f(x1), f(x2), …., f(xn)
corresponding to the values x = x 0, x1, x2, …, xn where the values of x are not necessarily
equispaced. From the definition of divided difference
f(x) - f(x0)
f[x,x0] = ----------------
x - x0
f(x) = f(x0) + (x-x0) f[x,x0] .... .... (1)
Again,
f[x,x0] - f[x0,x1]
f[x,x0,x1] = --------------------
x - x1

Page 13 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

f[x,x0] = f[x0,x1] + (x-x1) f[x,x0,x1] ... ...(2)

Substituting in (1) we get,
f(x) = f(x0) + (x-x0) f[x0,x1] + (x-x0) (x-x1) f[x,x0,x1]
Proceeding in this way we get,
f(x) = f(x0) + (x-x0) f[x0,x1] + (x-x0) (x-x1) f[x0,x1,x2]+... ...
+ (x-x0) (x-x1).. ... (x-xn-1) f[x0,x1,... ..,xn]
+ (x-x0) (x-x1).. ... (x-xn) f[x,x0,x1,... ..,xn]
If f(x) is a polynomial of degree n then, the (n+1)th order divided differences of f(x) will
be zero.
∴ f[x,x0,x1,... ...xn] = 0
Hence, f(x) = f(x0) + (x-x0) f[x0,x1] + (x-x0) (x-x1) f[x0,x1,x2]+... ...
+ (x-x0) (x-x1).. ... (x-xn-1) f[x0,x1,... ..,xn]
This is called Newton’s divided difference formula.
Example 3: Use Newton divided difference formula to find f(6) for the following
table.
X 5 7 11 13 21
f(x) 150 392 1452 2366 9702
Solution:
X f(x) ▲ ▲2 f(x) ▲3 f(x) ▲4 f(x)
f(x)
5 150
121
7 392 24
265 1
11 1452 32 0
457 1
13 2366 46
917
21 9702

We have f[x0] = 150, f[x0, x1] = 121, f[x0, x1, x2] = 24, f[x0, x1, x2, x3] = 1
f(6) = 150 + (6 - 5) (121) + (6 - 5) (6 - 7) (24) + (6 - 5) (6 - 7) (6 – 11) (1) + 0 + …
= 150 + 121 –24 + 5
= 252
Example 4: Page-104 [G. Shanker Rao].

Page 14 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Page 15 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

Inverse Interpolation
Given a set of values of x and y for y = f(x), the process of finding the value of x for a
certain value of y is called inverse interpolation.
Lagrange’s Inverse Interpolation:
We know, the Lagrange interpolation formula is
(x-x1) (x-x2) ... ... (x-xn) (x-x0) (x-x2) ... ... (x-xn)
y = f(x) = -------------------------------- f(x0) + ------------------------------------ f(x1)
(x0-x1) (x0-x2) ... ... (x0-xn) (x1-x0) (x1-x2) ... ... (x1-xn)

(x-x0) (x-x1) ... ... (x-xn-1)

+ ... ... + ---------------------------------- f(xn)
(xn-x0) (xn-x1) ... ... (xn-xn-1)
By interchanging x and y we can express x as a function of y as follows:
(y-y1) (y-y2) ... ... (y-yn) (y-y0) (y-y2) ... ... (y-yn)
x = -------------------------------- x0 + ------------------------------------ x1
(y0-y1) (y0-y2) ... ... (y0-yn) (y1-y0) (y1-y2) ... ... (y1-yn)

(y-y0) (y-y1) ... ... (y-yn-1)

+ ... ... + ---------------------------------- xn
(yn-y0) (yn-y1) ... ... (yn-yn-1)
This is called Lagrange’s inverse interpolation formula.
Example 7: Page-110 [G. Shanker Rao].

Method of Successive Approximations

Let the values of the independent variable x be given as x 0, x1, x2, …, xn where xi are
equispaced with h as the interval of differencing i.e. xi = x0 + ih, h = 0, 1, 2, …,n.
From Newton’s forward interpolation formula we have
y = y0 + u∆y0 + u(u-1)(Δ2y0)/ 2! + ... ... + u(u-1)(u-2) ... ...(u – n +1) (Δny0)/n!
where, u = (x – x0) / h
From this we obtain,
u = 1/Δy0 (yu - y0 - u(u-1)(Δ2y0)/ 2! - ... ... - u(u-1)(u-2) ... ...(u – n +1) (Δny0)/n!)
Neglecting the second and higher order differences, we obtain the first approximation to
u and this we write as follows
u1 = 1/Δy0 (yu - y0)
Now, we obtain the second approximation of u by including the term containing the
second differences. Thus .
u2 = 1/Δy0 (yu - y0 – u1 (u1-1)(Δ2y0)/ 2!)
where we have used the value of u1 for u in the co-efficient of Δ2y0.
Similarly we obtain the values of u3 and so on.
un = 1/Δy0(yu - y0 – un-1(un-1-1)(Δ2y0)/ 2! - ... ... ...- un-1 (un-1-1)( un-1-2) ... ...
... ( un-1 – n +1) (Δny0)/n!)
This process should be continued till two successive approximations to u agree with each
other to the required accuracy.

Page 16 of 17
MSA / L-4 CSE-4745 - Numerical Methods February 1, 2021

If un denotes the interpolated value of u, then x = x0 + unh gives the required value of x
for a given value of y.

 Similarly, we use same technique with Newton’s backward interpolation formula and
Central difference interpolation formula and interpolate x for a given y.
Example: Tabulate y = x3 for x = 2, 3, 4 and 5 and calculate the cube root of 10 correct to
3 decimal places.
Solution:
x y Δ Δ2 Δ3
2 8
19
3 27 18
37 6
4 64 24
61
5 125

Here yu = 10, y0 = 8, Δy0 = 19, Δ2 y0 = 18 and Δ3y0 = 6 and h = 1.

The successive approximations to u are therefore
u1= 1/19 (10 – 8 ) = 0.1053
u2 = 1/19 (2 – ( (0.1053(0.1053 – 1))/2 )).18 = 0.1499
u3 = 0.1532
u4 = 0.1541
u5 = 0.1542
∴ un = 0.154
Hence, x = x0 + unh = 2 + 0.154 = 2.154

Page 17 of 17