Numerical Methods & Analysis
Post [Link]. – 2026
What is this about ??
What are the issues ??
Numerical Methods and Analysis
What is this about & Why should we learn it ??
Numerical analysis:
• It is an area of mathematics and computer science that creates and implements
algorithms for obtaining numerical solutions to problems that are otherwise
difficult to solve in real life (no analytic or “closed form” solution / time-consuming
analytic solution while the error of numerical approximation is acceptable / involve
continuous variables / evaluation of experimental data) !!
• Analysis and mathematical modelling of the problems are the two key aspects.
• Such problems arise throughout the natural sciences, social sciences, engineering,
medicine, and business.
Numerical methods:
• These are the methods followed to carry out numerical analysis and modelling of
the problems !!
Advantage:
• It enables more realistic models to be treated.
Jan - Mar 2026 Post [Link].- 2026 2
Numerical Methods and Analysis
How do the numerical and analytical methods compare ??
Numerical Analytical
A close approximation An exact solution
Valid only for a defined interval (time, Valid at any interval (time, distance)
distance)
Applicable to many systems Applicable to one system
Large systems with many interactions Small closed systems
Small budget (use of computers saves Large budget (time, and manpower
time, manpower, money) required to construct a solution)
Lesson: Judgement of a problem from various viewpoints, like, the accuracy,
computation time and resources, budgets etc. is the foremost task before seeking its
solution.
Jan - Mar 2026 Post [Link].- 2026 3
Numerical Methods and Analysis
Numerical simulation in detector instrumentation !!
Physical processes inside a RPC detector
-6 kV
Bakelite Geometry
Gas Bakelite = 2 mm
Gas = 2 mm
Bakelite
+6 kV
Electrostatic field, E = V/d = 12 kV/6 mm = 20 kV/cm
Jan - Mar 2026 Post [Link].- 2026 4
Numerical Methods and Analysis
Numerical simulation in detector instrumentation !!
What happens in this geometry??
✓ Bakelite is resistive
✓ Finite dimension
✓ Spacer components
Electrostatic field inside RPC for 6 kV (BEM calculation)
Jan - Mar 2026 Post [Link].- 2026 5
Numerical Methods and Analysis
An example of dynamic process !!
Avalanche event Streamer event
41 kV/cm 47 kV/cm
Growth of electron density in RPC (FEM calculation)
Jan - Mar 2026 Post [Link].- 2026 6
Numerical Methods and Analysis
Numerical methods !!
Operations that numerical analysis performs (i.e. numerical methods)
• Find where 𝑓 𝑥 = 0 for a nonlinear equation or system of equations.
• Solve systems of linear equations, even large systems.
• Interpolate to find intermediate values from a table of values and fit curves to
experimental data.
• Approximate functions with polynomials or with a ratio of polynomials.
• Approximate values for the derivatives of a function, even if this is known only
by a table of function values.
• Evaluate the definite integral for any integrand, even if its values are known only
from experimental observations.
• Solve differential equations when initial values are given; these can be of any
order and complexity. Numerical analysis can even solve them if conditions are
specified at the boundaries of a region.
• Find the minima or maxima of functions, even when subject to constraints.
• Solve all types of partial differential equations by several techniques.
Jan - Mar 2026 Post [Link].- 2026 7
Numerical Methods and Analysis
Issues with numerical analysis !!
Issue # 1: Order and rate of convergence [valid for iterative methods]
• It describes how quickly a sequence of approximations reaches very close to
the solution.
• An iterative method is called convergent if the corresponding sequence
converges for given initial approximations.
Issue # 2 : Accuracy
• It concerns the difference between the exact solution and the approximate
one. It refers to how closely a computed or measured value agrees with the
true value.
Issue # 3 : Stability
• An algorithm is called numerically stable if an error, whatever its cause, does
not grow much larger during the calculation.
Jan - Mar 2026 Post [Link].- 2026 8
Numerical Methods and Analysis
Order and rate of convergence !!
It is expected that a convergent iteration in the general form of 𝑥𝑛+1 = 𝑔(𝑥𝑛 ) will
eventually converge to the true solution 𝑥 ∗ of the problem at the limit when 𝑛 → ∞.
The following expression represents how quickly the error converges to 0.
𝑒𝑛+1 𝑥𝑛+1 − 𝑥 ∗
lim = lim =𝜇
𝑛→∞ 𝑒𝑛 𝑝 𝑛→∞ 𝑥𝑛 − 𝑥 ∗ 𝑝
Here 𝑝 ≥ 1 is called the order of convergence and the constant 𝜇 is the rate of
convergence or asymptotic error constant.
Consider the following cases:
• If 𝑝 = 1 and 𝜇 = 1, the convergence is sublinear
• If 𝑝 = 1 and 0 < 𝜇 < 1, the convergence is linear
• If 𝑝 = 1 and 𝜇 = 0, the convergence is superlinear
• If 𝑝 = 2 and 𝜇 > 0, the convergence is quadratic
• If 𝑝 = 3 and 𝜇 > 0, the convergence is cubic
Jan - Mar 2026 Post [Link].- 2026 9
Numerical Methods and Analysis
Examples of convergence !!
Linear convergence: 𝑝𝑛 ≈ 0.5 𝑝𝑛−1 ≈ 0.5 2 𝑝𝑛−2 ≈ ⋯ ≈ (0.5)𝑛 𝑝0
𝑛 2𝑛
Quadratic convergence: 𝑝𝑛 ≈ 0.5 𝑝𝑛−1 2 ≈ 0.5 0.5 𝑝𝑛−2 2 2 … ≈ (0.5)2 −1 𝑝0
Lesson: A sequence with higher order of convergence converges more rapidly than a
lower order one.
Jan - Mar 2026 Post [Link].- 2026 10
Numerical Methods and Analysis
Accuracy (Errors in numerical analysis) !!
Modelling errors
Mathematical modelling is a process to represent physical reality by mathematical
equations which itself may be erroneous in describing the problem.
Example:
• The simplest model for population growth: 𝑁 𝑡 = 𝑁0 𝑒 𝑘𝑡
where 𝑁 𝑡 = population at time 𝑡, and 𝑁0 and 𝑘 are positive constants.
[For some stages of population growth, when it has unlimited resources, this can be
an accurate model. But more often, it overestimates the actual population for large
time.]
• SIR model in describing the spread of infectious disease (rubella measles)
𝑑𝑠(𝑡) 𝑑𝑖 𝑡 𝑑𝑟(𝑡)
= −𝑎 𝑠 𝑡 𝑖 𝑡 =𝑎𝑠 𝑡 𝑖 𝑡 −𝑏𝑖 𝑡 = 𝑏 𝑖(𝑡)
𝑑𝑡 𝑑𝑡 𝑑𝑡
6.8 1
where 𝑠 𝑡 = susceptible, 𝑖 𝑡 = infected, 𝑟 𝑡 = removed and 𝑎 = , 𝑏 =
11 11
[The same model works for other diseases, such as flu, and covid too, with a
suitable change in constants a and b. Again this is an approximation of the reality.]
Jan - Mar 2026 Post [Link].- 2026 11
Numerical Methods and Analysis
Errors in numerical analysis !!
Human-made errors
These can happen always and may arise from two following sources.
• Blunder / mistake
It might involve arithmetic error or programming error.
• Physical measurement error
✓ Calculation based on the physical data will contain the effect of observational
error.
✓ Numerical analysis cannot remove it, but can look at its propagated effect.
✓ It can suggest the best form for a calculation that will minimize the effect of
propagated errors on the data.
Jan - Mar 2026 Post [Link].- 2026 12
Numerical Methods and Analysis
Errors in numerical analysis !!
Machine representation related errors
✓ These are inevitable in computation where floating point representation is
involved.
✓ These may arise from chopping/truncation or rounding of floating point
numbers.
Problem: Computational difficulties arise from using a finite collection of machine
number to represent all the real numbers.
Representation of machine in the normalized decimal floating-point form:
±0. 𝑑1 𝑑2 … 𝑑𝑘 × 10𝑛
where 1 ≤ 𝑑1 ≤ 9, 0 ≤ 𝑑𝑖 ≤ 9 for each 𝑖 = 2, … , 𝑘
[k-digit decimal machine number]
A positive real number: 𝑥 = 0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛
The floating-point form of 𝑥, denoted by 𝑓𝑙(𝑥), for k-digit decimal machine
number representation is obtained by performing termination, called
chopping/truncation and rounding.
Jan - Mar 2026 Post [Link].- 2026 13
Numerical Methods and Analysis
Errors in numerical analysis !!
Chopping/Truncation
It simply chops off the digits 𝑑𝑘+1 𝑑𝑘+2 . . . . This produces the floating-point form:
𝑓𝑙(𝑥) = 0. 𝑑1 𝑑2 … 𝑑𝑘 × 10𝑛
Rounding
It adds 5 × 10−(𝑘+1) to 𝑥 and then chops the result to obtain a number of the
form: 𝑓𝑙(𝑥) = 0. 𝛿1 𝛿2 … 𝛿𝑘 × 10𝑛
➢ Rounding-up: Add 1 to 𝑑𝑘 to obtain 𝑓𝑙 𝑥 when 𝑑𝑘+1 ≥ 5 and then 𝛿𝑖 ≠ 𝑑𝑖 and
the exponent 𝑛 might change.
➢ Rounding-down: Simple chopping of 𝑑𝑘+1 onwards to obtain 𝑓𝑙 𝑥 when
𝑑𝑘+1 < 5 and then 𝛿𝑖 = 𝑑𝑖 for each 𝑖 = 2, … , 𝑘
Example: 5-digit a) chopping and b) rounding values of irrational number 𝜋 =
3.14159265. . . .
a) The floating-point form of π using five-digit chopping: 𝑓𝑙(𝜋) = 0.31415 × 101
b) The floating-point form of π using five-digit rounding: 𝑓𝑙(𝜋) = 0.31416 × 101
Jan - Mar 2026 Post [Link].- 2026 14
Numerical Methods and Analysis
Absolute and relative errors !!
Absolute error = 𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒 − 𝐴𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑖𝑜𝑛
𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒 −𝐴𝑝𝑝𝑟𝑜𝑥𝑖𝑚𝑎𝑡𝑖𝑜𝑛
Relative error = 𝑇𝑟𝑢𝑒 𝑣𝑎𝑙𝑢𝑒
Example: Absolute and relative errors when 𝑝 is approximated by 𝑝∗
𝑝 𝑝∗ Absolute error Relative error
0.3000 × 101 0.3100 × 101 0.1 0.3333ത × 10−1
0.3000 × 10−3 0.3100 × 10−3 0.1 × 10−4 0.3333ത × 10−1
0.3000 × 104 0.3100 × 104 0.1 × 103 0.3333ത × 10−1
Lesson: As a measure of accuracy, the absolute error can be misleading and the
relative error more meaningful, because the relative error takes into consideration
the size of the value.
Jan - Mar 2026 Post [Link].- 2026 15
Numerical Methods and Analysis
Errors in numerical analysis !!
Chopping and rounding errors
The floating-point representation 𝑓𝑙(𝑥) for the number 𝑥 has the relative error:
𝑥 − 𝑓𝑙(𝑥)
𝑥
Relative error for chopping of positive real number 𝑥 for k decimal digit floating-
point form 𝑓𝑙 𝑥 is:
0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛 − 0. 𝑑1 𝑑2 … 𝑑𝑘 × 10𝑛
0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛
0. 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛−𝑘 0. 𝑑𝑘+1 𝑑𝑘+2 … −𝑘
= = × 10
0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛 0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 …
Since 𝑑1 ≠ 0, the minimal denominator value is 0.1.
The numerator is bounded above by 1. Therefore, the relative error for chopping is:
𝑥 − 𝑓𝑙(𝑥) 1
≤ × 10−𝑘 = 10−𝑘+1
𝑥 0.1
Jan - Mar 2026 Post [Link].- 2026 16
Numerical Methods and Analysis
Errors in numerical analysis !!
Chopping and rounding errors
The floating-point representation 𝑓𝑙(𝑥) for the number 𝑥 has the relative error:
𝑥 − 𝑓𝑙(𝑥)
𝑥
Relative error for rounding of positive real number 𝑥 for k decimal digit floating-
point form 𝑓𝑙 𝑥 is (when 𝑑𝑘+1 < 5):
0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛 − 0. 𝑑1 𝑑2 … 𝑑𝑘 × 10𝑛
0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛
0. 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛−𝑘 0. 𝑑𝑘+1 𝑑𝑘+2 … −𝑘
= = × 10
0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 … × 10𝑛 0. 𝑑1 𝑑2 … 𝑑𝑘 𝑑𝑘+1 𝑑𝑘+2 …
When 𝑑𝑘+1 ≥ 5, then
𝑓𝑙 𝑥 = 0. 𝑑1 𝑑2 … 𝑑𝑘 × 10𝑛 + 10−𝑘
𝑥−𝑓𝑙(𝑥)
Home work: Show that ≤ 0.5 × 10−𝑘+1
𝑥
Jan - Mar 2026 Post [Link].- 2026 17
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
Assume that the floating-point representations 𝑓𝑙(x) and 𝑓𝑙(𝑦) are given for the
real numbers 𝑥 and 𝑦. We will assume a finite-digit arithmetic is given by:
𝑥 + 𝑦 = 𝑓𝑙 𝑓𝑙 𝑥 + 𝑓𝑙 𝑦
𝑥 − 𝑦 = 𝑓𝑙 𝑓𝑙 𝑥 − 𝑓𝑙 𝑦
𝑥 × 𝑦 = 𝑓𝑙 𝑓𝑙 𝑥 × 𝑓𝑙 𝑦
𝑥 ÷ 𝑦 = 𝑓𝑙(𝑓𝑙 𝑥 ÷ 𝑓𝑙 𝑦 )
This arithmetic corresponds to performing exact arithmetic on the floating-point
representations of 𝑥 and 𝑦 and then converting the exact result to its finite-digit
floating-point representation.
Rounding arithmetic is easily implemented while implementing finite-digit
chopping arithmetic is more difficult and requires a sequence of steps or a
procedure.
Jan - Mar 2026 Post [Link].- 2026 18
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
5 1
Example 1: Suppose that 𝑥 = 7 and 𝑦 = 3. Use five-digit chopping for calculating
𝑥 + 𝑦, 𝑥 − 𝑦, 𝑥 × 𝑦, 𝑥 ÷ 𝑦.
5 1
Given, 𝑥 = 7 = 0.714285ത and 𝑦 = 3 = 0. 3ത
5
Five-digit chopping : 𝑓𝑙(𝑥) = 7 = 0.71428 × 100 , 𝑓𝑙(𝑦) = 0.33333 × 100
𝑥 + 𝑦 = 𝑓𝑙 1.04761 × 100 = 0.10476 × 101
5 1 22
True value: 𝑥 + 𝑦 = 7 + 3 = 21
22
Absolute error: − 0.10476 × 101 = 0.190 × 10−4
21
Relative error:
Operation Result Actual value Absolute error Relative error
0.190 × 10−4 𝑥+𝑦 0.10476 × 101 22Τ21 0.190 × 10−4 0.182 × 10−4
22/21 𝑥−𝑦 0.38095 × 100 8Τ21 0.238 × 10−5 0.625 × 10−5
−4
𝑥×𝑦 0.23809 × 100 5Τ21 0.524 × 10−5 0.220 × 10−4
= 0.182 × 10
𝑥÷𝑦 0.21428 × 101 15Τ7 0.571 × 10−4 0.267 × 10−4
The arithmetic produces satisfactory five-digit results.
Jan - Mar 2026 Post [Link].- 2026 19
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
5 1
Example 2: Suppose, in addition to 𝑥 = 7 and 𝑦 = 3, we have 𝑢 = 0.714251, v =
98765.9, w = 0.111111 × 10−4
Five-digit chopping : 𝑓𝑙(𝑢) = 0.71425 × 100 , 𝑓𝑙(𝑣) = 0.98765 × 105 , and
𝑓𝑙 w = 0.11111 × 10−4
Determine 𝑥 − 𝑢, 𝑥 − 𝑢 ÷ 𝑤, 𝑥 − 𝑢 × 𝑣, 𝑢 + 𝑣
Operation Result Actual value Absolute error Relative error
𝑥−𝑢 0.30000 × 10−4 0.34714 × 10−4 0.471 × 10−5 0.136
(𝑥 − u) ÷ 𝑤 0.27000 × 101 0.31242 × 101 0.424 0.136
(𝑥 − 𝑢) × 𝑣 0.29629 × 101 0.34285 × 101 0.465 0.136
𝑢+𝑣 0.98765 × 105 0.98766 × 105 0.161 × 101 0.163 × 10−4
Operation 1: A small absolute error, but a large relative error (x and u are close).
Operation 2 & 3: Subsequent division by small number w or multiplication by large
number v magnifies the absolute error without modifying the relative error.
Operation 4: The addition of the large and small numbers u and v produces large
absolute error but not large relative error.
Jan - Mar 2026 Post [Link].- 2026 20
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
One of the most common error-producing calculations involves the
cancelation/loss of significant digits due to the subtraction of nearly equal
numbers.
5 5
Example: Check the previous case where 𝑥 = 7 and 𝑓𝑙(𝑥) = 7 = 0.71428 × 100 ,
𝑢 = 0.714251 and 𝑓𝑙(𝑢) = 0.71425 × 100
Then, 𝑓𝑙 𝑥 − 𝑓𝑙(𝑢) = 0.00003 × 100 and 𝑓𝑙 𝑓𝑙 𝑥 − 𝑓𝑙 𝑢 = 0.3? ? ? ?× 10−4
➢ However, in most calculation devices, 𝑥 − 𝑦 operation will be assigned the
machine specific significant digits.
➢ But any further calculations involving 𝑥 − 𝑦 retain the problem of loss of digits
of significance, since a chain of calculations is no more accurate than its weakest
portion.
Home work: Let 𝑝 = 0.54617 and 𝑞 = 0.54601. Use four-digit arithmetic to
approximate 𝑝 − 𝑞 and determine the absolute and relative errors using (a)
rounding and (b) chopping.
Jan - Mar 2026 Post [Link].- 2026 21
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
If a finite-digit representation or calculation introduces an error, further
enlargement of the error occurs when dividing by a number with small magnitude
(or, equivalently, when multiplying by a number with large magnitude).
Example: Check the previous case where 𝑥 − 𝑢 operation has small absolute error
0.471 × 10−5 and large relative error 0.136.
When divided by small number w = 0.111111 × 10−4 (or multiplied by a large
number v = 98765.9), the absolute error enlarges by an order of 105 and no
change in relative error is observed.
Operation Result Actual value Absolute error Relative error
𝑥−𝑢 0.30000 × 10−4 0.34714 × 10−4 0.471 × 10−5 0.136
(𝑥 − u) ÷ 𝑤 0.27000 × 101 0.31242 × 101 0.424 0.136
(𝑥 − 𝑢) × 𝑣 0.29629 × 101 0.34285 × 101 0.465 0.136
𝑢+𝑣 0.98765 × 105 0.98766 × 105 0.161 × 101 0.163 × 10−4
Jan - Mar 2026 Post [Link].- 2026 22
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
The loss of accuracy due to round-off error can often be avoided by a
reformulation of the calculations.
Example: Find the roots of 𝑥 2 + 62.10𝑥 + 1 = 0 using four-digit rounding
arithmetic whose roots are 𝑥1 = −0.01610723 and 𝑥2 = −62.08390
Since 𝑏2 ≫ 4𝑎𝑐, the numerator of 𝑥1 involves subtraction of nearly equal number.
𝑏 2 − 4𝑎𝑐 = (62.10)2 − 4.000 1.000 1.000 = 3856. −4.000 = 3852. = 62.06
−62.10+62.06 −0.04000
𝑓𝑙 𝑥1 = = = −0.02000 (subtraction of nearly equal numbers)
2.000 2.000
−0.01611+0.02000 −1
Relative error: ≈ 2.4 × 10
−0.01611
−62.10−62.06 −124.02
𝑓𝑙 𝑥2 = = = −62.10 (addition of nearly equal numbers)
2.000 2.000
−62.08+62.10 −4
Relative error: ≈ 3.2 × 10
−62.08
Poor approximation of 𝑥1 may be addressed by changing the form of quadratic
formula.
Jan - Mar 2026 Post [Link].- 2026 23
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
The loss of accuracy due to round-off error can often be avoided by a
reformulation of the calculations.
Example: Find the roots of 𝑥 2 + 62.10𝑥 + 1 = 0 using four-digit rounding
arithmetic whose roots are 𝑥1 = −0.01610723 and 𝑥2 = −62.08390 (contd.)
Change the quadratic formula to determine 𝑥1 through rationalization of the
numerator:
−𝑏 + 𝑏2 − 4𝑎𝑐 −𝑏 − 𝑏2 − 4𝑎𝑐 𝑏2 − (𝑏2 − 4𝑎𝑐)
𝑥1 = =
2𝑎 2
−𝑏 − 𝑏 − 4𝑎𝑐 2𝑎(−𝑏 − 𝑏2 − 4𝑎𝑐)
−2𝑐
𝑥1 =
𝑏 + 𝑏2 − 4𝑎𝑐
−2(1.000) −2.000
𝑓𝑙 𝑥1 = = = −0.01610
62.10 + 62.06 124.2
−0.01611+0.01610
Relative error: ≈ 6.2 × 10−4
−0.01611
Jan - Mar 2026 Post [Link].- 2026 24
Numerical Methods and Analysis
Errors in numerical analysis !!
Finite digit arithmetic errors
The loss of accuracy due to round-off error can often be avoided by a
reformulation of the calculations.
Example: Find the roots of 𝑥 2 + 62.10𝑥 + 1 = 0 using four-digit rounding
arithmetic whose roots are 𝑥1 = −0.01610723 and 𝑥2 = −62.08390 (contd.)
What happens if the rationalization technique is applied for 𝑥2 ?
−2𝑐 −2(1.000) −2.000
𝑓𝑙 𝑥2 = = = = −50.00
2
−𝑏 − 𝑏 − 4𝑎𝑐 62.10 − 62.06 0.0400
−62.08+50.00
Relative error: ≈ 1.9 × 10−1
−62.08
It leads to not only the subtraction of nearly equal numbers, but also the division
by the small result of this subtraction which finally produces quite a large relative
error.
This technique may be useful when 𝑏 is negative.
Lesson: Think before you compute!
Jan - Mar 2026 Post [Link].- 2026 25
Numerical Methods and Analysis
Stability !!
➢ Accumulation of errors may lead to numerical instability.
➢ An algorithm is called numerically stable if an error, whatever its cause, does
not grow much larger during the calculation.
• Numerical stability can be attained if the problem is well-conditioned meaning
that the solution changes by only a small amount if the problem data are
changed by a small amount. To the contrary, if a problem is ill-conditioned, then
any small error in the data will grow to be a large error.
• Both the original problem and the algorithm can be well-conditioned and/or ill-
conditioned and any combination is possible.
• An art of numerical analysis is to find a stable algorithm for solving a well-posed
problem.
Jan - Mar 2026 Post [Link].- 2026 26
Numerical Methods and Analysis
An example of numerical stability !!
Computation of the value of 2 (~ 1.41421) [a well-posed problem]
Algorithm
• Start with an initial approximation 𝑥0 (say 1.4)
• Compute improved guesses 𝑥1 , 𝑥2 , . . etc.
𝑥𝑛 1
Method 1 : 𝑥𝑛+1 = + 𝑥 [Babylonian method]
2 𝑛
Method 2 : 𝑥𝑛+1 = (𝑥𝑛 − 2)2 +𝑥𝑛
2
Initial guesses : 𝑥0 = 1.4 and 𝑥0 = 1.42
Method 2 Method 2
Method 1 Method 1
𝑥0 = 1.4 𝑥0 = 1.42
𝑥0 = 1.4 𝑥0 = 1.42 𝑥1 = 1.4016 … 𝑥1 = 1.42026896 …
𝑥1 = 1.4142857 … 𝑥1 = 1.41422535 … 𝑥2 = 1.4028614 … 𝑥2 = 1.42056 …
𝑥2 = 1.414213564 … 𝑥2 = 1.41421356242 … ..... .....
𝑥1000000 = 1.41421 … 𝑥27 = 7280.2284 …
Jan - Mar 2026 Post [Link].- 2026 27