lOMoARcPSD|31568146

papers for all ai

Artificial Intelligence/Expert Systems In Design And Manufacturing (SRM Institute of

Science and Technology)

Scan to open on Studocu

Studocu is not sponsored or endorsed by any college or university

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)
 lOMoARcPSD|31568146

SRM Institute of Science and Technology

College of Engineering and Technology Mode of Exam
School of Computing OFFLINE
SRM Nagar, Kattankulathur – 603203,
Chengalpattu District, Tamilnadu
Academic Year:2024-2025 (ODD) SET D

Test: CT-1 Date: 02.09.2024

Course Code & Title: 21CSC305P- MACHINE LEARNING Duration: 1 hr 40
mins
Year & Sem: III & V Max. Marks: 50

Course Articulation Matrix:

S.No. Course Outcome PO1 PO PO PO PO PO PO PO PO PO1 PO11 PO1
2 3 4 5 6 7 8 9 0 2

1 CO1 H H - H - - - - - - - -

2 CO2 M H - H - - - - - - - -

Part - A
( 10 x 1 = 10 Marks)
Instructions: Answer all
Q. Question
No
1 You created machine learning system that interacts with its environment and responds to errors and
rewards. What type of machine learning system is it?
a) Supervised learning
b) Reinforcement learning
c) Semi-supervised learning
d) Unsupervised learning
2 In a dataset, if the variance is zero, what does that indicate about the data points?
a) All data points are identical.
b) All data points are different.
c) The data points follow a normal distribution.
d) The data points are evenly spaced.
3 If X and Y are NOT independent, which among the following is correct?
a) Var(X+Y)=Var(X)+Var(Y)
b) Var(X+Y)=Var(X)+Var(Y)+Cov(X,Y)
c) Var(X+Y)=Var(X)+Var(Y)-Cov(X,Y)
d) Var(X+Y)=Var(X)+Var(Y)+2Cov(X,Y)
4 Method in which the previously calculated probabilities are revised with values of new probability is called
__________
a) Sum rule
b) Bayes theorem
c) Dependent theorem
d) Product rule
5 There is a class of 40 students out of which 16 are girls. There are 27 students who are right-handed. How
many minimum numbers of girls who are left-handed in this class?
a) 17
b) 56
c) 23
d) 3

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)

 lOMoARcPSD|31568146

6 The best fit line method for data in Linear Regression?

a) Least Square Error
b) Maximum Likelihood
c) Logarithmic Loss
d) Both A and B
7 Why might a probabilistic discriminative model outperform a generative model in a classification task?
a) It always uses more complex algorithms.
b) It directly models the conditional probability P(Y∣X) and focuses on classification boundaries.
c) It can generate new data samples.
d) It typically assumes fewer dependencies between features.
8 Which of the below models is a generative model used in machine learning?
a) Support vector machines
b) Bayesian
c) Logistic Regression
d) Linear Regression
9 Previous probabilities in Bayes Theorem that are changed with help of new available information are
classified as _________________
a) independent probabilities
b) posterior probabilities
c) interior probabilities
d) dependent probabilities
10 What do you mean by a hard margin?
a) The SVM allows a very low error in classification
b) The SVM allows a high amount of error in the classification
c) Number of cross-validations
d) Trade off in kernel

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)

 lOMoARcPSD|31568146

Part – B
(4 x 5 = 20Marks)
Instructions: Answer any 4
11 Explain the process of discovering clusters in unsupervised learning. Mention the applications of clustering
method.
ANSWER:
Our first goal is to estimate the distribution over the number of clusters, p(K|D); this tells us if there are
subpopulations within the data. For simplicity, we often approximate the distribution p(K|D) by its mode, K∗ =
arg maxK p(K|D). In the supervised case, we were told that there are two classes (male and female), but in the
unsupervised case, we are free to choose as many or few clusters as we like. Picking a model of the “right”
complexity is called model selection, and will be discussed in detail below
Our second goal is to estimate which cluster each point belongs to. Let zi ∈ {1,...,K} represent the cluster to
which data point i is assigned. latent variable, since it is never observed in the training set.) We can infer which
cluster each data point belongs to by computing z∗ i = argmaxk p(zi = k|xi, D). This is illustrated in Figure
1.8(b), where we use different colors to indicate the assignments, assuming K = 2.
Here are some real world applications of clustering.
• In astronomy, the autoclass system (Cheeseman et al. 1988) discovered a new type of star, based on clustering
astrophysical measurements.
• In e-commerce, it is common to cluster users into groups, based on their purchasing or web-surfing behavior,
and then to send customized targeted advertising to each group (see e.g., (Berkhin 2006)).
• In biology, it is common to cluster flow-cytometry data into groups, to discover different sub-populations of
cells
12 Discuss in detail about any two PDF in probability and statistics to model different types of data.

The Normal distribution is one of the most important PDFs in probability and statistics due to its wide
applicability and the central role it plays in the Central Limit Theorem. It is defined by two parameters: the
mean (μ) and the standard deviation (σ). The PDF of a normal distribution is given by:

The Exponential distribution is another key PDF that models the time between events in a
Poisson process, where events occur continuously and independently at a constant average
rate. It is defined by one parameter, the rate parameter (λ), and its PDF is given by:

13 Find the Variance value for the continuous random variable, ( ) = x2 , for the interval 1<= x<= 2

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)

 lOMoARcPSD|31568146

14 You are given the following data points for x and Y as (2,3), (4,6), (6,9), (8,12), (10, 15). Using the least
squares method, find the equation of the line y=mx +b that best fits the given data points.

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)

 lOMoARcPSD|31568146

15 You are training three different models on a dataset. After evaluating each model, you obtain the following
errors:
Model A: Training Error: 1.0, Test Error: 10.0
Model B: Training Error: 5.0, Test Error: 5.5
Model C: Training Error: 6.0, Test Error: 6.0
Tasks: Identify which model is likely overfitting, underfitting, and best fitting. Briefly explain why you classify
each model as overfitting, underfitting, or best fitting with necessary graph.
Solution:
Model A: Training Error: 1.0, Test Error: 10.0 : Overfitting
Model A has a very low training error but a significantly higher test error. This indicates that the model
performs exceptionally well on the training data but fails to generalize to new, unseen data, suggesting
overfitting.
Model B: Training Error: 5.0, Test Error: 5.5: Best fitting
Model B shows similar training and test errors. The errors are relatively close and not excessively high,
indicating that the model generalizes well and strikes a good balance between fitting the training data and
performing on unseen data.
Model C: Training Error: 6.0, Test Error: 6.0 Underfitting
Model C has both training and test errors that are relatively high and equal. This suggests that the model is too
simple to capture the underlying patterns in the data, leading to poor performance on both training and test
data, indicating underfitting.

16 Explain in detail about Hinge loss function in SVM

Part – C
(2x 10 = 20 Marks)
17 Explain the concept of polynomial curve fitting. How does the order of the polynomial affect the fitting
process, and what are the potential consequences of choosing a polynomial that is too low or too high in order?
Answer:
• Probability theory provides a framework for expressing such uncertainty in a precise and quantitative
manner.

where M is the order of the polynomial. The polynomial coefficients w0,...,wM are collectively denoted by the
vector w.
• The error function is given by the sum of the squares of the errors between the predictions y(xn, w) for
each data point xn and the corresponding target values tn, so that we minimize

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)

 lOMoARcPSD|31568146

Effect of Polynomial Order:

Low Order Polynomial:
Characteristics: A polynomial of low order (e.g., linear or quadratic) provides a simple model that might not
capture the complexity of the data.
Consequences:
 Underfitting: The model might not capture the underlying trends in the data, resulting in high bias and
poor performance on both training and test data.
 Example: Fitting a linear polynomial to a dataset with a non-linear relationship might lead to significant
errors.
High Order Polynomial:
 Characteristics: A polynomial of high order (e.g., quartic or higher) can fit the data points more closely,
potentially capturing intricate patterns.
Consequences:
 Overfitting: The model might fit the training data extremely well, including noise and outliers, which can
lead to poor generalization to new, unseen data. This results in high variance.
 Example: Fitting a 10th-degree polynomial to a few data points might produce a curve that oscillates
wildly and does not generalize well.
Graphical Representation:
 Low Order Polynomial: A straight line or simple curve that might not capture the complexity.
 High Order Polynomial: A more complex curve that fits the training data points closely but may
fluctuate.
(OR)
b) A company manufactures two types of products: Product X and Product Y. The proportions of products are
as follows:
 30% of the products are Product X.
 70% of the products are Product Y.
The company uses a quality control test to determine if a product is defective. The defect rates for each product
are:
 The probability that Product X is defective is 8%.
 The probability that Product Y is defective is 12%.
If a randomly selected product is found to be defective, what is the probability that it is Product X?

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)

 lOMoARcPSD|31568146

18 a) What is ridge linear regression, and how does it differ from traditional linear regression in handling
dependent and independent variables? Describe the approaches used in robust linear regression to minimize the
impact of datapoints on the estimated relationship between variables.
Answer:
One problem with ML estimation is that it can result in overfitting. Ridge regression- way to ameliorate this
problem by using MAP estimation with a Gaussian prior.

(OR)
b) How does the two class and multiclass criterion optimize the separation between classes in a linear
classification model, and how does this contribute to dimensionality reduction?
Answer:

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)

 lOMoARcPSD|31568146

Course Outcome (CO) and Bloom’s level (BL) Coverage in Questions

Approved by the Audit Professor/Course Coordinator

Downloaded by Charmi Bandapalli (charmibandapalli@gmail.com)