Numeric computatioN

In the field of numeric computation, computers perform

mathematical operations on real-world data. However,
because computers represent numbers in a finite manner,
special techniques are used to handle issues such as
floating-point arithmetic, errors, and iterative methods
for solving equations. This unit will cover the basic
concepts of floating-point numbers, their operations, error
analysis, and some important iterative methods for solving
equations numerically.
1. Computer Arithmetic and Floating Point Numbers
Floating-point arithmetic is used to represent real numbers
in a way that can accommodate a wide range of values by
using a combination of a mantissa and an exponent. This
representation allows for both very small and very large
numbers, but it also introduces the possibility of errors due
to the finite precision of the system.
1.1 Floating-Point Representation
A floating-point number is generally represented as:

For example, the number 123.456 can be represented in

scientific notation as:

In binary (base-2), this becomes:

This is how floating-point numbers are represented in

computers, using a sign bit, exponent, and mantissa. A
floating-point number is limited by the number of bits
used to store the exponent and mantissa.
1.2 Floating Point Operations
The operations on floating-point numbers, such as addition,
subtraction, multiplication, and division, are carried out by
performing these operations on the mantissas, adjusting
the exponent accordingly.
However, because floating-point numbers have limited
precision, errors may arise during these operations due to
rounding. For instance, small discrepancies can occur
when a result is truncated or rounded to fit within the
available precision of the computer.
1.3 Normalization
Normalization is the process of adjusting the representation
of floating-point numbers so that the mantissa falls within
a certain range. In the case of binary floating-point
numbers, this typically means ensuring that the first digit
of the mantissa is 1 (for base-2). This process maximizes
the precision of the floating-point representation and
minimizes errors.
For example:
• The number 0.00123 can be normalized as
in scientific notation.
• Similarly, in binary, would be normalized to

Normalization helps in ensuring that calculations are as

accurate as possible given the constraints of the computer
system.
1.4 Consequences of Floating-Point Representation
Floating-point representation, while allowing for a wide range
of values, has its drawbacks:
 • Rounding Errors: Since the mantissa has finite precision,
the result of an arithmetic operation may differ slightly
from the exact result.
• Overflow and Underflow: Overflow occurs when the
result of an operation exceeds the maximum
representable number. Underflow occurs when the
result is smaller than the smallest representable
number.

2. Errors in Numeric Computation

In numeric computation, errors can arise in various ways,
especially when approximating real-world numbers. These
errors are categorized into absolute error, relative error,
and percent error.

2.1 Absolute Error

The absolute error is the difference between the exact value
and the computed (or approximated) value.
Mathematically:
For example, if the true value is 5.567 and the computed
value is 5.56, the absolute error is:

2.2 Relative Error

Relative error is the absolute error divided by the true value.
It provides a measure of how large the error is in relation
to the size of the number:

For example, if the true value is 5.567 and the approximate

value is 5.56:

2.3 Percent Error

Percent error is simply the relative error multiplied by 100 to
express it as a percentage:

For the same example:

These errors are important in numerical analysis because
they help determine the accuracy of the results of
calculations.

3. Iterative Methods for Solving Equations

When solving transcendental or polynomial equations
numerically, iterative methods are used because analytical
solutions are often not available or are difficult to
compute. These methods are based on successive
approximations to reach a solution.
3.1 Zeros of Transcendental Equations
A transcendental equation is one that cannot be solved by
algebraic means and involves transcendental functions like
trigonometric, exponential, and logarithmic functions. For
example, the equation may not have a closed-
form solution, but iterative methods can approximate the
solution.
3.2 Bisection Method
The bisection method is a simple and reliable root-finding
method that works on continuous functions. It is based on
the Intermediate Value Theorem, which states that if a
continuous function has opposite signs at two points
‘a’ and ‘b’, then there must be ‘a’ root between a and ‘b’.
The method proceeds by:

1. Finding the midpoint

 2. Evaluating
3. Replacing either ‘a’ or ‘b’ with ‘c’, depending on the sign
of and repeating until the solution converges.
This method guarantees convergence, but it may take many
iterations if the interval is large.

3.3 False Position Method

The false position method (also known as the Regula Falsi
method) is similar to the bisection method but uses linear
interpolation to find the root. In this method:
1. We choose two initial points ‘a’ and ‘b’ where and
have opposite signs.
2. A straight line is drawn between these two points.
3. The intersection of this line with the x-axis is used as the
new approximation for the root.
 4. The process is repeated by replacing one of the initial
points with the new approximation.
The false position method can converge faster than the
bisection method but still guarantees convergence.

3.4 Newton-Raphson Method

The Newton-Raphson method is an iterative method that
uses the derivative of the function to approximate the
root. The method starts with an initial guess and
iteratively refines it using the formula:

Here:
• is the function value at
• is the derivative of the function at
• is the new approximation.
This method is often faster than the bisection or false
position methods but requires that the derivative of the
function be known and non-zero at the root.

3.5 Convergence of Solution

The convergence of these iterative methods refers to how
quickly the solution approaches the true root. For
example:
• The bisection method always converges, but it may
require many iterations.
• The false position method typically converges faster
than bisection.
• The Newton-Raphson method has quadratic
convergence (i.e., the number of correct digits doubles
with each iteration), which makes it the fastest among
the three methods, provided a good initial guess is
chosen.
Example
Let’s consider an example equation:

We want to find the root using the bisection method.

1. Choose an initial interval, say

2. Compute the midpoint

3. Evaluate is the root.
This method converges quickly in this case, but if the initial
guess were further from the root, more iterations would
be required.

Conclusion
Numerical computation plays a critical role in the field of
computer science and engineering, providing the tools and
techniques to solve complex mathematical problems.
Floating-point arithmetic, error analysis, and iterative
methods are foundational for numerical methods, and
understanding these concepts is essential for handling
real-world computations efficiently. The methods
discussed, such as bisection, false position, and Newton-
Raphson, are indispensable for solving equations that
cannot be easily tackled algebraically. By understanding
how these methods work and their convergence
properties, we can ensure more accurate and efficient
solutions in a wide range of applications.