Download as pdf or txt
Download as pdf or txt
You are on page 1of 35

NPTEL Online Certification Courses

Indian Institute of Technology Kharagpur

Introduction to Soft Computing


Assignment 3
TYPE OF QUESTION: MCQ
Number of questions: 10 Total mark: 10 X 2 = 20

Question 1: Which fuzzy inference system approach prioritizes interpretability at the expense of
accuracy?

a. Mamdani approach
b. Takagi and Sugeno's approach
c. Both approaches provide the same level of interpretability and accuracy.
d. The choice depends on the specific application and its priorities.

Correct Answer: a

Detailed Explanation: The Mamdani approach follows linguistic fuzzy modelling and is characterized by
its high interpretability but low accuracy. On the other hand, Takagi and Sugeno’s approach follows
precise fuzzy modelling and obtains high accuracy but at the cost of low interpretability.

Question 2: What does the term "linguistic hedges" refer to in a fuzzy logic system?

a. Techniques for defining the universe of discourse


b. Methods for calculating membership values
c. Modifiers that adjust the meaning of linguistic terms
d. Algorithms for fuzzy inference

Correct Answer: c

Detailed Explanation:

Linguistic hedges are modifiers that can be applied to linguistic terms to adjust their meaning or degree
of membership, providing a more nuanced representation in fuzzy logic systems.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Question 3: Consider a fuzzy logic system where the input variable is “Temperature” ranging from 0°C
to 50°C. The fuzzy sets for this variable are defined as “Cold”, “Warm”, and “Hot”. The membership
function for the “Warm” set is a triangular function defined by the points (10, 0), (25, 1), (40, 0),
where the function increases linearly from 10°C to 25°C and decreases linearly from 25°C to 40°C. If
the current temperature is 30°C, what is the membership value of the “Warm” set?

a. 0.33
b. 0.5
c. 0.75
d. 0.67

Correct Answer: d

Detailed Explanation:

The membership function for the “Warm” set is a triangular function defined by the points (10, 0), (25, 1),
and (40, 0). This means that the function increases linearly from 10°C to 25°C and decreases linearly from
25°C to 40°C.

Given that the current temperature is 30°C, we are in the decreasing phase of the “Warm” set. We can
calculate the membership value by finding the slope of the line between the points (25, 1) and (40, 0) and
using it to find the y-value at x=30.

(0−1) −1
The slope(𝑚) of the line is (40−25)
= 15
.

Now, plug the slope into the equation 𝑌 = 𝑚𝑋 + 𝑐 and solve for 𝑐 using the point (25,1):

−1
𝑐 =1−( × 25) = 8/3
15
−1 8
i.e. the equation of the line is: 𝑌 = 15
𝑋 +3

Now, To find the membership value at 30°C substitute 𝑋 = 30 into the equation.

−1 8 2
𝑌= × 30 + = = 0.6666
15 3 3

Question 4: Consider a fuzzy logic system that controls a heater in a room. The system takes two inputs:
the temperature x and the humidity y. The fuzzy sets for x are Cold and Hot with membership
functions μCold(x) and μHot(x), respectively. Similarly, the fuzzy sets for y are Dry and Wet with
membership functions μDry(y) and μWet(y), respectively. The output z controls the heater and has
Low and High fuzzy sets.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

The system has two rules:

1. If the room is Cold and Dry, then the heater is set to Low.
2. If the room is Hot and Wet, then the heater is set to High.

Now, suppose on a particular day, the membership functions for the inputs are as follows: μCold(x)
= 0.4, μHot(x) = 0.6, μDry(y) = 0.5, and μWet(y) = 0.5. Can you compute the rule
strengths for the two rules?

a. 0.5, 0.5
b. 0.2, 0.3
c. 0.4, 0.5
d. 0.24,0.25

Correct Answer: c

Detailed Explanation:

The rule strength (also called α values) of the firable rules is calculated as the minimum of the membership
functions of the inputs.

For Rule 1: α(R1) = min(μCold(x), μDry(y)) = min(0.4, 0.5) = 0.4

For Rule 2: α(R2) = min(μHot(x), μWet(y)) = min(0.6, 0.5) = 0.5

So, the rule strengths for Rule 1 and Rule 2 are 0.4 and 0.5, respectively.

Question 5: High interpretability is followed by the fuzzy logic controller (s)

a. Mamdani approach
b. Takagi and sugeno’s approach
c. Both a, b
d. None of the above

Correct Answer: a

Detailed Explanation:

High interpretability is typically associated with the Mamdani approach in fuzzy logic controllers. In the
Mamdani model, linguistic rules expressed in natural language are utilized, making it easier for humans to
comprehend and interpret the system's behavior.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Question 6: Limitations of the traditional optimization approaches are

a. Discrete (integer) variables are difficult to handle.


b. Method may not be suitable for parallel computing.
c. Methods may not necessarily be adaptive.
d. All the above.

Correct Answer: d

Detail explanation:

Traditional optimization approaches face several limitations. Firstly, they struggle with handling
discrete (integer) variables, posing challenges in scenarios where variables must take on non-
continuous values. Secondly, these methods may not be well-suited for parallel computing, limiting
their scalability and efficiency in modern computing environments. Lastly, traditional optimization
methods often lack adaptability, as they may have fixed strategies and parameters that cannot
dynamically adjust to changes in the optimization problem landscape. These limitations highlight the
need for more advanced and flexible optimization techniques in certain contexts.

Question 7: The genetic algorithm works based on the mechanisms of

a. Neural network and natural language processing


b. Natural genetics and natural evaluation.
c. Artificial intelligence and machine learning.
d. None of these.

Correct Answer: b

Detail explanation:

The genetic algorithm, rooted in natural genetics, mimics the process of natural selection and
evolution to solve optimization and search problems. Inspired by the principles of natural evolution,
the algorithm employs a population of potential solutions, representing candidate solutions as
individuals. Through the iterative process of selection, crossover, and mutation, individuals undergo
genetic operations to generate new offspring. The fitness of each solution is evaluated based on its
performance in solving the problem at hand. Over successive generations, solutions evolve, and the
algorithm converges towards optimal or near-optimal solutions. This approach draws parallels with
the way genetic traits propagate and adapt in biological systems, providing an effective and nature-
inspired optimization technique.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Question 8: Combinatorial optimization is


a. NP‐Hard problem
b. NP‐complete problem
c. Scholastic Assessment problem
d. None of the above

Correct Answer: a

Detail explanation:

This classification implies that finding an optimal solution in polynomial time is deemed
computationally infeasible, particularly as the problem size increases. NP-hard problems, including
combinatorial optimization, do not possess known efficient algorithms for finding the global
optimum. The difficulty arises from the exponential growth in the number of possible solutions that
need to be explored. Combinatorial optimization involves selecting the best solution from a finite set
of feasible solutions, and its computational complexity contributes to its categorization as an NP-hard
problem. Although heuristics and approximation algorithms are commonly employed to address
combinatorial optimization challenges, there is no known polynomial-time algorithm that
guarantees the optimal solution for all instances of these problems.

Question 9: In Steady-state Genetic algorithm

a. Population size is small and Chromosomes are of longer length.


b. Population size is small and Chromosomes are of smaller length.
c. Population size is large and Chromosomes are of smaller length.
d. Population size is large and Chromosomes are of longer length.

Correct Answer: a

Detail explanation:

Population size is small, and chromosomes are of longer length. This configuration is chosen to
maintain a smaller population, promoting a focus on the most promising solutions. Meanwhile,
having longer chromosomes allows for a more extensive representation of potential solutions, which
can be beneficial in certain optimization scenarios. The small population size aids in intensifying the
search within a limited set of solutions, emphasizing exploitation over exploration.

Question 10: Fuzzy Logic Controller contains


a. Fuzzy rule base
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

b. Fuzzy inference engine


c. A fuzzification module
d. All the above

Correct Answer: d

Detail explanation:

A Fuzzy Logic Controller (FLC) integrates a fuzzy rule base, fuzzy inference engine, and fuzzification
module. The fuzzy rule base encapsulates expert knowledge using linguistic variables and terms,
establishing relationships between input and output. The fuzzy inference engine processes these rules,
combining fuzzy information to generate an output. The fuzzification module converts crisp input values
into fuzzy sets, employing linguistic terms and membership functions. Collectively, these components
enable FLCs to model and control systems in the presence of uncertainty, imprecision, and vagueness,
making them effective in various applications where traditional control approaches may fall short.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Introduction to Soft Computing


Assignment 4
TYPE OF QUESTION: MCQ
Number of questions: 10 Total mark: 10 X 2 = 20

Question 1: Which encoding method is useful in ordering problems such as the Travelling Salesman
Problem (TSP)?

a. Binary Encoding
b. Permutation Encoding
c. Value Encoding
d. None of the above

Correct Answer: b

Detailed Explanation: In permutation encoding, each solution is represented as a permutation of the


numbers from 1 to n, where n is the number of cities (or nodes) in the TSP. The order of the numbers in
the permutation corresponds to the order in which the cities are visited. This encoding ensures that each
city is visited exactly once in a given solution.

Question 2: Using a binary encoding scheme for a job-shop scheduling problem with three jobs (J1, J2,
and J3) and two machines (M1 and M2), how would you represent a solution where J1 is scheduled on
M1, J2 is scheduled on M2, and J3 is not scheduled on any machine?

a. 100100
b. 100000
c. 100110
d. 101100

Correct Answer: a

Detailed Explanation:

10: This represents J1, with the first bit (1) indicating it's scheduled on M1 and the second bit (0)
indicating it's not scheduled on M2.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

01: This represents J2, with the first bit (0) indicating it's not scheduled on M1 and the second
bit (1) indicating it's scheduled on M2.

00: This represents J3, with both bits (0) indicating it's not scheduled on either machine.

Question 3: Consider a binary tree-encoded genetic algorithm. The tree represents arithmetic
expressions where the nodes can be operators (+, -, *, /). If the maximum depth of the tree is 3 (root
is at depth 0), what is the maximum number of nodes the tree can have?

a. 15
b. 7
c. 45
d. 9

Correct Answer: a

Detailed Explanation:

The maximum number of nodes at a depth 𝑑 in a binary tree is 2𝑑 . So, for a tree with a maximum depth
of 3, the total number of nodes is 20 + 21 + 22 + 23 = 1 + 2 + 4 + 8 = 15.

Question 4: In the context of genetic algorithms and their selection processes, why might the Roulette
Wheel Selection method be considered advantageous yet potentially problematic?

a. It guarantees the selection of the best individual in every generation, ensuring rapid
convergence.
b. It distributes selection probabilities based on fitness, favouring fitter individuals but
potentially sidelining lesser-fit individuals over time.
c. It always promotes diversity by selecting individuals randomly, ensuring a wide range of
solutions in subsequent generations.
d. It eliminates the need for a fitness function, relying solely on random chance for selection.

Correct Answer: b

Detailed Explanation:

Roulette Wheel Selection allocates selection probabilities based on an individual's fitness: fitter
individuals have a higher chance of being selected. While this can lead to faster convergence towards
optimal solutions, it also poses a risk. Over successive generations, the method may consistently favour
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

the same set of top-performing individuals, reducing genetic diversity and potentially missing out on
exploring other promising regions of the solution space. Thus, while advantageous in prioritizing fitter
solutions, it requires careful handling to prevent premature convergence and maintain genetic diversity.

Question 5: Which type of generation gap typically leads to higher selection pressure?

a. No gap
b. Small gap
c. Large gap
d. The type of gap doesn't affect selection pressure

Correct Answer: c

Detailed Explanation:

In a large generation gap, most of the offspring replace the parents. This means that only the fittest
individuals from the parent generation are likely to contribute to the offspring, leading to higher
selection pressure.

Question 6: In real value encoding, if XL =2 and XU = 17 are two extreme decoded values of a variable
x. n= 4 is the number of binary bits in the representation of x. If the decoded value of x is 10, then x
will be
a. 1100
b. 1010
c. 1110
d. 1111

Correct Answer: a

Detailed explanation:
𝑋𝑈 −𝑋𝐿
𝑋 = 𝑋𝐿 + . 𝑋𝐵
2𝑛 −1
17−2
= 2 + 24 −1 . 10
15
= 2 + 15 . 10

= 12
The binary of 12 is (1100)2
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Question 7: If the mating pool size is N, how many mating pairs are to be formed in Random Mating?
𝑁
a.
4
𝑁
b. 2
𝑁2
c. 2
𝑁2
d. 4

Correct Answer: b

Detailed explanation:

In genetic algorithms, for Random Mating, mating pairs are typically formed from the mating pool,
𝑁
and each pair contributes to the creation of offspring. Therefore, if the mating pool size is N, then 2
mating pairs are formed. Each pair undergoes crossover or recombination to produce new
individuals in the next generation.

Question 8: The generation gap with N population size and p number of individuals is
𝑝
a. 𝑁
𝑃2
b. 𝑁
𝑝
c. 𝑁2
2𝑁
d. 𝑝

Correct Answer: a

Detailed explanation:

The generation gap is defined as the proportion of individuals in the population who are replaced in
each generation.

Question 9: Which coding scheme has evolved to overcome the binary code-related problem?

a. Pink coding.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

b. Grey coding.
c. Yellow coding.
d. Black coding.

Correct Answer: b

Detailed explanation:

Grey coding is a binary numeral system in which two consecutive numbers differ in only one bit. This
coding scheme is designed to reduce the possibility of errors during transitions between consecutive
numbers, as only one bit changes at a time. In contrast, traditional binary coding can lead to multiple
bits changing simultaneously, which can result in errors or glitches, especially in digital systems. Grey
coding helps in minimizing errors and simplifying transitions in various applications, making it
particularly useful in contexts where accurate and reliable encoding and decoding are essential.

Question 10: Which is not the drawback of the binary crossover problem?

a. End point Bias.


b. Hamming cliff problem.
c. Positional bias.
d. Uniform variation.

Correct Answer: d

Detailed explanation:

The other options (Endpoint Bias, hamming cliff problem, Positional bias) are considered drawbacks
associated with certain crossover techniques, but "Uniform variation" is not typically considered a
drawback in the context of binary crossover problems. Uniform variation refers to a type of mutation
operator where each bit in the chromosome has an equal chance of being flipped or changed. It is a
form of introducing diversity in the population and can be a useful exploration mechanism.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Introduction to Soft Computing


Assignment 5
TYPE OF QUESTION: MCQ
Number of questions: 10 Total mark: 10 X 2 = 20

Question 1: In blend crossover, what is the purpose of the alpha and beta parameters?

a. They control the spread of children's solutions


b. They determine the length of the chromosome
c. They represent the gene values of parents
d. They define the crossover probability

Correct Answer: a

L
Detailed Explanation: Alpha and beta specify the "spread" of possible values for children's genes. Alpha

E
defines the base interval around the midpoint between the parents' values, while beta allows for further

T
expansion of that interval in one direction (usually towards the better parent). Higher alpha and beta
values lead to a wider range of possible solutions in the offspring, promoting exploration, while lower

P
values result in offspring closer to the parents, favoring exploitation.

N
Question 2: How does simulated binary crossover differ from linear and blend crossover in
terms of choosing alpha values?

a. Simulated binary crossover requires a probabilistic approach for alpha


b. Linear crossover uses a probability distribution function for alpha
c. Blend crossover uses a constant value for alpha
d. Simulated binary crossover guarantees global optima

Correct Answer: a

Detailed Explanation: Unlike linear and blend crossovers, which use fixed or adjustable alpha values,
Simulated binary crossover employs a probability distribution function (PDF) to draw alpha values for each
crossover operation.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Question 3: Consider two parent sequences for a genetic algorithm:

Parent 1: [1, 2, 3, 4, 5, 6, 7, 8] Parent 2: [8, 7, 6, 5, 4, 3, 2, 1]

What offspring sequences will be if we perform a single-point order crossover after position 4?

a. Offspring 1: [1, 2, 3, 4, 5, 6, 7, 8], Offspring 2: [8, 7, 6, 5, 1, 2, 3, 4]


b. Offspring 1: [1, 2, 3, 4, 8, 7, 6, 5], Offspring 2: [8, 7, 6, 5, 1, 2, 3, 4]
c. Offspring 1: [1, 2, 3, 4, 4, 3, 2, 1], Offspring 2: [8, 7, 6, 5, 5, 6, 7, 8]
d. Offspring 1: [1, 2, 3, 4, 1, 2, 3, 4], Offspring 2: [8, 7, 6, 5, 8, 7, 6, 5]
Correct Answer: b

Detailed Explanation: For a thorough breakdown of the algorithm, check the details provided on slide
31.

EL
T
Question 4: Consider two parent sequences for a genetic algorithm: Parent 1: [A, B, C, D, E, F, G, H, I]
Parent 2: [I, H, G, F, E, D, C, B, A]. What are the offspring sequences after performing the position-based

P
crossover operation with the positions 3, 5, and 7?

N
a. Offspring 1: [I, B, G, D, E, F, C, H, A], Offspring 2: [A, H, C, F, E, D, G, B, I]
b. Offspring 1: [I, H, C, F, E, D, G, B, A], Offspring 2: [A, B, G, D, E, F, C, H, I]
c. Offspring 1: [A, B, G, D, E, F, C, H, I], Offspring 2: [I, H, C, F, E, D, G, B, A]
d. Offspring 1: [A, H, C, F, E, D, G, B, I], Offspring 2: [I, B, G, D, E, F, C, H, A]

Correct Answer: b

In the position-based crossover operation, the positions selected in the first parent should be copied to
the first child, and the remaining alphabets should be filled from the second parent in their order, starting
from the first position. The same process should be repeated for the second child, but starting from the
second parent.

Question 5: When might the inversion operation be performed in a genetic algorithm?

a. After each crossover


b. At the beginning of the algorithm
c. Only in the final iteration
d. Occasionally, on selected chromosomes
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Correct Answer: d

Detailed Explanation: This operation is not performed after each crossover, at the beginning of the
algorithm, or only in the final iteration. Instead, it’s applied occasionally on selected chromosomes to
introduce diversity in the population and avoid premature convergence.

Question 6: The fitness scaling is useful to avoid

a. premature convergence, and slow finishing.


b. mature convergence, and fast finishing
c. premature convergence, and fast finishing
d. mature convergence, and slow finishing

Correct Answer: a

Detailed Explanation:

EL
T
Fitness scaling is used to avoid premature convergence, where the algorithm settles on a
suboptimal solution too quickly, and to promote a more thorough search for an optimal solution,

P
leading to slower finishing. It helps prevent the algorithm from getting stuck in local optima by

N
providing a balance between exploration and exploitation.

Question 7: In the context of Multi-Objective Optimization Problems (MOOPs), a set of trade-


off optimal solutions is commonly known as:
a. Optimal Solutions
b. Non-Dominated Solutions
c. Dominated Solutions
d. Converged Solutions
Correct Answer: b

Detailed Explanation: In Multi-Objective Optimization Problems (MOOPs), the goal is to find


solutions that optimize multiple conflicting objectives simultaneously. A set of trade-off optimal
solutions is commonly referred to as Pareto optimal solutions. These solutions are non-
dominated, meaning that no other solution in the set can improve one objective without
worsening at least one other objective. In other words, Pareto optimal solutions represent a set
of solutions where no single solution is superior to another in all objectives. The concept is
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

derived from Pareto efficiency, named after the economist Vilfredo Pareto, and it reflects the
idea of achieving the best possible outcomes without making any other outcome worse.

Question 8: In Multi-Objective Optimization, a Utopian objective vector corresponding to a


solution represents:

a. A solution with the worst possible objective values


b. A solution with objective values is strictly better than any in the search space
c. A solution with objective values equal to the best in the search space
d. A solution with randomly assigned objective values

Correct Answer: b

L
Detailed Explanation: A Utopian objective vector corresponds to an ideal or utopian solution, and

E
in the context of multi-objective optimization, it represents a solution that has objective values
strictly better (improving in all objectives) than any other solution in the entire search space. This

T
implies that the utopian solution is the best possible solution achievable across all objectives.

P
Question 9: Suppose the original solution is 15.6, the random number lying between 0.0 to 1.0 is 0.7,

N
and the maximum value of the perturbation decided by the user is 2.5. In Random mutation, the
mutation value will be

a. 15.6
b. 17.7
c. 16.1
d. 23.4
Correct Answer: b

Detailed Explanation:

𝑃𝑃𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 = 15.6, 𝑟𝑟 = 0.7, 𝛥𝛥 = 2.5

𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 = 15.6 + (0.7 − 0.5) ∗ 2.5 = 16.1


NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Question 10: In Multi-Objective Optimization, the Nadir objective vector corresponds to:

a. A solution with the worst possible objective values


b. A solution with objective values strictly better than any in the search space
c. A solution with objective values equal to the best in the search space
d. A solution with randomly assigned objective values

Correct Answer: a

Detailed Explanation:

The Nadir objective vector represents a solution in multi-objective optimization with the worst
possible objective values. It serves as a reference point for the worst achievable values in each

L
objective within the search space. This helps in understanding the trade-offs and performance of

E
other solutions by comparing their objective values against the Nadir point. Therefore, option (a)

T
is the correct answer.

N P
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Introduction to Soft Computing


Assignment 6
TYPE OF QUESTION: MCQ
Number of questions: 10 Total mark: 10 X 2 = 20

Question 1: What property does the domination relation in a multi-objective optimization problem
exhibit?

a. Reflexive
b. Symmetric

L
c. Transitive
d. Antisymmetric

E
Correct Answer: C

PT
Detailed Explanation:

The domination relation in multi-objective optimization is not reflexive, symmetric, or antisymmetric.

N
However, it is transitive, meaning if solution X dominates solution Y and solution Y dominates solution Z,
then solution X dominates solution Z.

Question 2: Consider two solutions, P and Q, for a problem where we want to minimize two objectives.
The solutions with their objective values are: 𝑃𝑃 = (1, 2) , 𝑄𝑄 = (2, 3) Which of the following are True?

a. P dominates Q.
b. Q dominate P.
c. Both dominate each other.
d. None of the above

Correct Answer: a

Detailed Explanation:

A solution dominates another solution if it is no worse on both objectives and strictly better on at least
one objective.

Given the solutions P = (1, 2) and Q = (2, 3), and considering that we want to minimize the objectives, we
can say:
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

● P dominates Q: This is True. P is better in both objectives (1 < 2 and 2 < 3).
● Q dominates P: This is False. Q is worse in both objectives (2 > 1 and 3 > 2).

So, the correct answer is “P dominates Q”.

Question 3: Which of the following statements accurately describes the relationship between non-
dominance sets and Pareto optimal sets in multi-objective optimization?

a. The non-dominance and Pareto optimal sets are the same and can be used interchangeably.
b. The non-dominance set is a strict subset of the Pareto optimal set.
c. The Pareto optimal set is a strict subset of the non-dominance set.

L
d. The non-dominance set and the Pareto optimal set are unrelated concepts.

E
Correct Answer: a

T
Detailed Explanation: In the context of multi-objective optimization, a solution is said to be Pareto

P
optimal if another solution that dominates it does not exist. In other words, a Pareto optimal solution is
not dominated by any other solution. Therefore, the set of all Pareto optimal solutions is often called

N
the non-dominance set. Hence, these two terms describe the same concept and can be used
interchangeably.

Question 4: Which of the following is NOT included in the list of non-Pareto-based multi-objective
optimization approaches?

a. Lexicographic ordering
b. Criterion selection (VEGA)
c. Simple weighted approach (SOEA)
d. Non-dominated Sorting based approach (NSGA)

Correct Answer: d

Please refer to slides, page number 30.

Question 5: What does the term "middling performance" refer to in the context of the VEGA approach?

a. Excellent performance in all objectives


b. Poor performance in all objectives
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

c. Results neither very bad nor very good


d. Global optimization

Correct Answer: c

Detailed Explanation: Please refer to slide number 60 in the lecture materials..

Question 6: What is the basis of NPGA (Niched Pareto Genetic Algorithm) in its selection scheme?

a. Roulette wheel selection

L
b. Tournament selection and Pareto dominance.
c. Stochastic universal sampling

E
d. Rank-based selection

T
Correct Answer: b

P
Detailed Explanation: The basis of NPGA (Non-dominated Sorting Genetic Algorithm) in its selection

N
scheme involves a combination of tournament selection and Pareto dominance. In multi-objective
optimization, NPGA uses Pareto dominance to identify nondominated individuals. Tournament selection
is then applied to choose individuals based on their Pareto ranks, promoting diversity and maintaining a
balanced representation of nondominated solutions in the population. This combination helps NPGA
efficiently explore the Pareto front and guide the evolutionary process toward finding a diverse set of
solutions.

_____________________________________________________________________________________

Question 7: How does NPGA determine the winner solution in its tournament selection?

a. The solution with the highest fitness value wins.


b. The solution with the lowest fitness value wins.
c. The solution based on Pareto dominance principle wins.
d. The solution selected randomly wins.

Correct Answer: c

Detailed Explanation: Explanation: In NPGA, the winner's solution in the tournament is determined using
the Pareto dominance principle, which ensures that the selected solution is non-dominated by others in
the comparison set.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

____________________________________________________________________________

Question 8: What is the lexicographic ordering method in multi-objective optimization?

a. An a priori technique employing stochastic sampling.


b. A posteriori technique based on dominance relationships.
c. An a priori technique based on the principle of "aggregation with ordering."
d. A selection strategy involving tournament-based competition.

E L
Correct Answer: c

T
Detailed Explanation: The lexicographic ordering method in multi-objective optimization is an a
priori technique that relies on "aggregation with ordering." In this method, objectives are ranked

P
based on their importance, and solutions are sorted lexicographically. The lexicographic ordering
considers the most important objective first, and if there are ties, it moves on to the next objective

N
in the order of importance. This process continues until a unique order is established for each
solution. Lexicographic ordering provides a deterministic way of comparing and selecting solutions
in multi-objective optimization based on their lexicographical order of objectives.

Question 9: In the context of Multi-Objective Genetic Algorithms (MOGA), what does "fitness sharing"
refer to?

a. A technique for evaluating the fitness of individuals based on their Pareto dominance.
b. A mechanism for sharing the fitness values equally among all individuals in the
population.
c. An approach for preventing premature convergence by penalizing solutions that are too
similar.
d. A method for randomly selecting individuals for crossover and mutation operations.

Correct Answer: c
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Detailed Explanation: Dominance-based ranking is a strategy in multi-objective optimization


algorithms where individuals in a population are ranked according to their Pareto dominance
relationships. Each individual's domination count, representing how many other individuals it
dominates, is determined by comparing objective values. Individuals that dominate no others are
assigned rank 1, forming the first non-dominated front. Subsequent fronts are created by
iteratively assigning ranks to remaining individuals. This approach organizes the population into
Pareto fronts, helping maintain diversity and guide evolutionary algorithms toward discovering
optimal trade-off solutions across conflicting objectives. It is a fundamental concept in algorithms
like NSGA-II and MOGA, facilitating effective selection and convergence towards Pareto fronts.

______________________________________________________________________________

E L
Question 10: What is a significant disadvantage of the Vector-Evaluated Genetic Algorithm (VEGA)
regarding solution dominance?

T
a. Solutions are locally and globally dominated

P
b. Solutions are Pareto-dominated and globally dominated
c. Solutions are locally non-dominated but not necessarily globally dominated.

N
d. Solutions are locally and globally non-dominated

Correct Answer: c

Detailed Explanation: The disadvantage of the Vector-Evaluated Genetic Algorithm (VEGA) is that the
solutions it generates are locally non-dominated, but their non-dominance is limited to the current
population and may not extend to being globally non-dominated across the entire search space.
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Introduction to Soft Computing


Assignment 7
TYPE OF QUESTION: MCQ
Number of questions: 10 Total mark: 10 X 2 = 20

Question 1: What is the time complexity of the Non-dominated Sorting Procedure in the context of
Genetic Algorithms, where m is the number of objectives, and n is the population size?

a. 𝑂𝑂(𝑛𝑛^2)
b. 𝑂𝑂(𝑚𝑚𝑚𝑚)

L
c. 𝑂𝑂(𝑚𝑚𝑚𝑚^2)
d. 𝑂𝑂(𝑙𝑙𝑙𝑙𝑙𝑙 𝑛𝑛)

E
Correct Answer: C

T
Detailed Explanation:

P
The time complexity of the Non-dominated Sorting Procedure in the context of Genetic Algorithms

N
(GAs), where m represents the number of objectives and n is the population size, is given by:

𝑂𝑂(𝑚𝑚𝑚𝑚^2)

For more explanation refer to slide number 8.

Question 2: What is the purpose of the sharing technique in NSGA?

a. To assign dummy fitness values


b. To increase computation speed
c. To favor elitism approach
d. To maintain population diversity

Correct Answer: d

Detailed Explanation:

The sharing technique promotes a diverse set of non-dominated solutions, ensuring that the algorithm
explores a wide range of possibilities and avoids getting stuck in local optima
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Question 3:What is the significance of the crowding distance method in NSGA-2?

a. It improves crossover operations


b. It ensures the fastest execution of the algorithm
c. It helps select diverse solutions from the last front
d. It enhances the mutation rate

Correct Answer: c

Detailed Explanation: The crowding distance method in NSGA-II is significant because it promotes
diversity among solutions in the same Pareto front. By measuring the distances between neighboring
solutions, it ensures that the population remains well-distributed, preventing clustering and encouraging

L
exploration of different regions of the Pareto front. In summary, it strikes a balance between
convergence and diversity, enhancing the quality of non-dominated solutions in multi-objective

E
optimization

a.
b.
c.
d.
Rank Selection

N PT
Question 4: Which selection method does NSGA-II use in its algorithm?

a. Roulette Wheel Selection

Crowded Tournament Selection


Random Selection

Correct Answer: c

NSGA-II uses Crowded Tournament Selection, refer page 38.

Question 5: What does the crowding comparison operator consider when selecting a winner between
two solutions in NSGA-II?

a. Rank of the solutions


b. Mutation rate
c. Crowding distance and rank
d. Crossover probability

Correct Answer: c
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Detailed Explanation: The crowding comparison operator in NSGA-II considers both crowding distance
and rank to determine the winner.

Question 6: The behavior of a biological neural network can be captured by a simple model called an
artificial neuron or perceptron. Which of the following statements about the perceptron is correct?

a. Perceptrons can only be used for linearly separable problems.


b. Perceptrons are a type of unsupervised learning algorithm.
c. Perceptrons have no activation function.
d. Perceptrons can have multiple layers, just like deep neural networks.

E L
Correct Answer: a

T
Detail explanation: Perceptrons are the building blocks of artificial neural networks and were initially
designed to address linearly separable problems. They have a binary output based on a weighted sum of

P
inputs, and their learning algorithm adjusts the weights to achieve correct classifications. However,
perceptrons have limitations, as they cannot handle non-linearly separable problems. To address this

N
limitation, more complex neural network architectures with activation functions and multiple layers, such
as multilayer perceptrons, were developed.

Question 7: In the context of neural network architectures, what distinguishes recurrent neural
networks (RNNs) from feedforward network architectures?

a. RNNs have no hidden layers.


b. RNNs have at least one "feedback loop."
c. RNNs only process input data once.
d. RNNs do not use activation functions.

Correct Answer: b

Detail explanation: Recurrent neural networks (RNNs) are characterized by the presence of at least one
feedback loop. Unlike feedforward neural networks, RNNs allow information to be circulated within the
network, enabling them to handle sequential and temporal data. The feedback loops introduce a form of
memory to the network, allowing it to consider previous inputs in the context of current inputs. This ability
makes RNNs suitable for tasks like natural language processing and time-series prediction. Options A, C,
and D are incorrect as they do not accurately describe the distinctive features of recurrent neural
networks.

Question 8: To address non-linearly separable problems, which type of neural network architecture is
commonly employed?
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

a. Single-layer perceptron network


b. Multilayer feedforward neural network
c. Radial Basis Function network
d. Recurrent neural network

Correct Answer: b

Detail explanation: Multilayer feedforward neural networks are designed to handle non-linearly
separable problems. These networks consist of an input layer, one or more hidden layers, and an output
layer. The inclusion of hidden layers with activation functions enables the network to capture complex
relationships in the data, making it suitable for solving problems that are not linearly separable. Options
A, C, and D are not as well-suited for non-linearly separable problems. Single-layer perceptron networks

L
(option A) are limited to linearly separable problems, radial basis function networks (option C) have a
specific structure for radial basis functions, and recurrent neural networks (option D) are more suitable

E
for sequential and temporal data.

T
Question 9: What is the fundamental structural unit of a biological nervous system consisting of
interconnected processing units?

P
a. Synapse

N
b. Neuron
c. Axon
d. Dendrite

Correct Answer: b

Detail explanation: The basic structural and functional unit of a biological nervous system is the neuron.
Neurons are specialized cells that process and transmit information through electrical and chemical
signals. They have three main parts: dendrites (receive signals), a cell body (processes signals), and an
axon (transmits signals to other neurons). The interconnected network of neurons allows for complex
processing and communication within the nervous system. While synapses (option A) are crucial for
communication between neurons, and axons and dendrites (options C and D) are parts of the neuron, the
neuron itself is the primary unit responsible for information processing in the nervous system.

Question 10: What is the chemical present in each neuron that plays a crucial role in transmitting signals
between neurons?

a. Enzyme
b. Hormone
c. Neurotransmitter
d. Antigen
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Correct Answer: c

Detail explanation: Neurotransmitters are chemicals present in neurons that play a vital role in
transmitting signals across synapses, the junctions between neurons. When an electrical signal reaches
the end of a neuron's axon, neurotransmitters are released into the synapse. These chemicals then bind
to receptors on the adjacent neuron, transmitting the signal from one neuron to another. This process is
essential for communication within the nervous system. Options A (Enzyme), B (Hormone), and D
(Antigen) are not the primary chemicals involved in transmitting signals between neurons. Enzymes
typically catalyze biochemical reactions, hormones are chemical messengers produced by glands, and
antigens are substances that trigger an immune response.

TE L
N P
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Introduction to Soft Computing


Assignment 8
TYPE OF QUESTION: MCQ
Number of questions: 10 Total mark: 10 X 2 = 20

Question 1: What is the primary purpose of the weight matrix in a perceptron during the training
process?

a. To store input data


b. To define the threshold value

L
c. To calculate the error gradient
d. To adjust the connection strengths between neurons

Correct Answer: d

Detailed Explanation:

PTE
N
In a perceptron, which is a fundamental unit in neural networks, the weight matrix signifies the synaptic
weights or connection strengths between the input neurons and the output neuron. During training, the
perceptron learns by fine-tuning these weights based on the observed input-output pairs. The weight
matrix plays a crucial role in determining how much influence each input feature has on the final output
prediction. By iteratively adjusting these weights, the perceptron adapts to enhance its performance, with
the ultimate goal of minimizing prediction errors.

Question 2: In Hebbian learning, what does the weight adjustment depend on?

a. Correlation between input and output


b. The number of layers in the network
c. The learning rate
d. The transfer function used

Correct Answer: a

Detailed Explanation:

Correlation between input and outputs is the core principle of Hebbian learning, captured by the phrase
"cells that fire together, wire together." If an input neuron and an output neuron are both active (either
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

both positive or both negative), the weight between them increases. If they are oppositely active, the
weight decreases. This positive correlation drives the network to strengthen connections that contribute
to the desired output.

Question 3: What role does the activation function play in the back-propagation algorithm for training
multilayer feed-forward neural networks?

a. It defines the initial weights of the network.


b. It calculates the error between predicted and actual outputs.
c. It determines the learning rate during weight updates.
d. It introduces non-linearity to the network and aids in capturing complex patterns.

E L
Correct Answer: d

T
Detailed Explanation:

P
The activation function in a neural network introduces non-linearity to the model. This nonlinearity is
crucial for the network to learn and represent complex patterns in the data. In the context of the back-

N
propagation algorithm, the activation function is applied to the output of each neuron in the hidden
layers.

If a neural network only consisted of linear activation functions, no matter how many layers it had, the
overall network would still behave as a linear model. The introduction of non-linear activation functions
allows the network to capture and learn intricate patterns and relationships within the data, making it
capable of solving more complex problems.

Question 4: What role does the learning rate play in the Steepest Descent method?

a. It represents the error function


b. It determines the size of the incremental steps in updating parameters
c. It is the output of the input layer
d. It represents the difference between true output and observed output

Correct Answer: b

In Steepest Descent and other gradient-based optimization methods, the learning rate controls how much
the parameters are adjusted in each iteration based on the calculated gradient. A higher learning rate
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

leads to larger steps, potentially reaching the minimum faster but possibly overshooting or oscillating
around it. A lower learning rate takes smaller steps, ensuring convergence but potentially slowing down
the process

Question 5: What is the key difference between supervised learning and unsupervised learning in the
context of neural networks?

a. Supervised learning requires labeled output data, while unsupervised learning does not.
b. Unsupervised learning uses a teacher to guide the learning process.
c. Supervised learning does not involve input data, unlike unsupervised learning.
d. Unsupervised learning requires a predefined threshold value.

L
Correct Answer: a

E
Detailed Explanation: In supervised learning, the neural network is trained with labeled output data,

T
while unsupervised learning does not require labeled output data.

b.
c.
d.
FIS Variable block

N
Fuzzy Controller block
Rule Base block
P
Question 6: In the Fuzzy Toolbox Library in the Simulink environment, which block is responsible for
implementing the desirable rules based on a FIS variable produced by the FIS Editor?

a. Defuzzification block

Correct Answer: c

Detailed Explanation: The Fuzzy Controller block in the Fuzzy Toolbox Library in Simulink is designed to
implement the desirable rules based on a Fuzzy Inference System (FIS) variable produced by the FIS Editor.
The Fuzzy Controller block takes the input variables, applies the fuzzy logic rules defined in the FIS, and
produces output variables. This block is a key component for building fuzzy logic controllers in Simulink.
Options A (Defuzzification block), B (FIS Variable block), and D (Rule Base block) are not the specific blocks
responsible for implementing rules based on FIS variables in the Simulink environment.

_____________________________________________________________________________________

Question 7: In a Sequential Hybrid System, how are the technologies utilized?

a. Parallel execution
b. Hierarchical execution
NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

c. Pipelining fashion
d. Synchronous execution

Correct Answer: c

Detailed Explanation: In a Sequential Hybrid System, the technologies are used in a pipelining
fashion. This means that different technologies or processes are executed sequentially, with each
stage or component working on a specific aspect of the overall task. The output of one stage
becomes the input for the next stage, creating a pipeline of operations. This approach allows for
efficient and continuous processing, with each technology contributing to the overall
functionality of the system. Options A (Parallel execution), B (Hierarchical execution), and D

L
(Synchronous execution) do not specifically describe the sequential, pipelined nature of the
technologies in a Sequential Hybrid System.

E
______________________________________________________________________________

T
Question 8: In an Embedded Hybrid System, how are the participating technologies integrated?

P
a. Sequential execution

N
b. Isolated execution
c. Intertwined integration
d. Modular execution

Correct Answer: c

Detailed Explanation:

In an Embedded Hybrid System, the participating technologies are integrated in such a manner that they
appear intertwined. This integration is characterized by a close and interconnected collaboration between
different technologies within the system. Unlike a sequential or modular approach, where each
component operates independently or in a distinct order, an embedded hybrid system involves a more
cohesive and interwoven combination of technologies. This integration is designed to optimize
performance, efficiency, and collaboration among various system elements. Options A (Sequential
execution), B (Isolated execution), and D (Modular execution) do not capture the integrated and
intertwined nature of technologies in an Embedded Hybrid System.

Question 9: What is a characteristic of a fuzzy rule base in a fuzzy logic system?


NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

a. It only contains linguistic terms.


b. It is not involved in decision-making.
c. Rules are expressed in a binary format.
d. It defines relationships between input and output variables.

Correct Answer: d

Detailed Explanation:

In a fuzzy logic system, a fuzzy rule base is responsible for defining relationships between input and output
variables. Fuzzy rules capture the linguistic knowledge and express the fuzzy logic reasoning used in
decision-making. Each rule typically consists of antecedents (conditions based on input variables) and
consequents (output variables or actions). The rules in the fuzzy rule base guide the system in making

L
decisions or generating output based on the fuzzy input information. Options A, B, and C do not accurately

E
describe the role and characteristics of a fuzzy rule base in a fuzzy logic system.

_____________________________________________________________________________________

T
Question 10: Why are fuzzy logic and neural networks considered natural complementary tools in

P
building intelligent systems in the context of Neuro-Fuzzy Systems?

N
a. Fuzzy logic focuses on raw data, while neural networks handle linguistic information.
b. Fuzzy logic and neural networks both operate at a high level of abstraction.
c. Neural networks excel in linguistic reasoning, while fuzzy logic deals with raw data.
d. Neural networks handle raw data well, and fuzzy logic is effective in dealing with linguistic
information.

Correct Answer: d

Detailed Explanation:

In the context of Neuro-Fuzzy Systems, neural networks are well-suited for processing raw data and
capturing complex patterns, while fuzzy logic excels in handling linguistic information and reasoning at a
higher level. Combining these two tools leverages their complementary strengths, allowing for a more
comprehensive approach to intelligent system development. Option D accurately captures the synergy
between neural networks and fuzzy logic in Neuro-Fuzzy Systems. Options A, B, and C do not accurately
represent the natural complementarity described in the statement.

You might also like