Nirma University

Institute of Technologr
Semester End Examination (IR/R,R), ilcembe r -
M' Tech. in computer scienc"
2or9
Engineering, semester-I /
M' Tech. in computer scienc. T3 Engin"eering (t-Ns), semesterll
Td
M. Tech. in computer science and Engineiring /
toii" science), semester_l
3CS1 1 1 1 Applied Machine Learning

RollNo. Supervisor's initial

with date
Time: 3 Hours
Max. Marks : 100
Instructions: 1. Attempt all questions.
2. Figures to right indicate full marks.
3. Use section-wise separate answer book.
4. Draw neat sketches wherever necessary.

Section - I
Q-1.
[16]
A. In the given data Earns is the class value. Use Narve Bayes to
CO2BLs predict how much does the Individual earns with detaits [10]
1"ss, 1,
self-Emp, Asian, 45). Explain in detail what assumptions are ,r".d
to calculate the probabilities.
rlities.
working hours
Age Male? Work Class Race per week Earns
39 1 State Govt white 40 <=50K
50 L Self-Emp Black 13 <=50K
38 0 Private Asian 40 <=50K
53 1 Private Asian 40 <=50K
53 1 Self-Emp White 45 >50K
31 0 Private White 50 >50K
42 L Private Asian 40 >50K
37 1 Private Asian 80 >50K
32 1 State Govt Black 40 <=50K
38 0 Private Black 50 <=50K
B What is the need of scaling input feature before applying anjof the
C02BL5 machine learning algorithm? Give proper example for the sarne. t6l

Q-2
[16]
A Data for the hours spent in studies against the score of the
CO2BL3 students is given in the table [10]
a. Determine a linear regression model equation to represent
this data
b. Plot the equation.
c. If a student studied for 15 hours per day, based upon this
study, what would be the expected score?

P^ry--t/t
 i
i
I

3CSI l I I-APPLIED MACHINE LEARNING

I Hours -- Spent Score Hours
I
Studying per day Spent Score
Studying per day
4 390 1i
I
9
790
i
580 I 3s0
10 650
i
3 400
74 730 B 590
4 410 11 640
7 530 5 450
t2 600 6 520

Eorur, f,.'ffi*Xr1:o[:t*t'" hard clustering and sort clustering

'ildth to6]

OR
B Related to error there are_ two possibre
C02BL5 error. Define the terms and explain terms true error and test to6l
using
"* """*pf".
Q-3.
(18)
A Linear regression has two approaches
CO2BL5 of probrem solving, gradient
descent and normal equation. what t6]
is thi criteria for using one
method over the otheri use an example justify
B There is a scenario of a robot .r.rig"to.r.
to your answer.
CO3BL6 and put the ball in the basket. what will wrr"rr"tie robot has to go 16]
learning algorithm? Also write the approach be the right machine
components. in detail and its
C For the tree given berow, check whether
CO1BL3 pruning is possible at any t6l
of the node. Justify your answ.. *ith
p.op., *giment.
Dependent variable; pLAy

OR
C Accuracy may not be correct parameter
CO 1BL2 algorithm to another. Explain frt* ."".rracy for comparlng one t6l
may be not at all be
indicating- the goodness. How olher parameters
evaluate the performarce can be used to
more appropriately.

P<Y-a1't
 3CS 1 I 1 I-APPLIED MACFIINE LEARNING

Section - II
Q-4.
[16]
Answer the following:
A consider the Neural Network with initialized weights as in the
CO2BL3 Figure 1. Explain the network architecture knowiig that we are t10I
tq"g to distinguish between nails and screws for an example of
training tuples is as follows: T1 {0.6, O.1, nai}, T2 screw}.
Implement the forward p.op.gaiion of uee signats{O.2,0,3,
in the network
using T1 and T2 as input. calcutate the errorr and only write the
expression for updating the weights using gradient descent
algorithm. Let the learning rate 11 Ue b. t.

orltl?tat
taellrcir.y
jt.idden
srer4rorr,s
5

B Define the term

t6l
COIBL2 a. Precision and recall
b. Learning Rate
c. Deep Learning
Q-s.
[16]
A For the data given square and Dots are two classes. They are not
CO2BL3 [10]
linearly separable The transformation function to be used is
defined as below.

0fl,)=l(::x: ;ll: :::l:) irffirrpz

\/
t (::) orherwise
calculate the parameters for the support vector machine that can
classify the data. Classify the point l-t,Z\ using the SVM.
(Hint: the transformed values is in rinearly separable space)

? ^xp_zl ?
 3CSI I I I.APPLIED MACHINE LEARNING
B Take 10 points in two dimension having coordinate
CO2BL2 shown use k mea,s crustering to cluster values as t6l
them into two clusters
also trace the way^centroid getJupdated
after first iteration. (2,10),
(3,5), (2,2), (4,5), (2,6]), (10,15), (12,L4l,, (t4,t1),
(13,13), (15,12).
OR

B How do you differentiate between Information

CO2BL2 Gain, Gain Ratio
Ginny Index? what are the advantage of one over and t6]
the other?
Q-6.
[ 18]

A Given a three problems t6]

CO3BL4 1' Estimating the price of the computer based on RAM, HDD,
Number of cores, frequency of the processor-
2. Deciding the person i" u. i"""*cher or not based on color of
lhi.lt hair(short I longl, wearing glasses(yes/no1, age.
^ Deciding
3. to admit tiie studerr-t l, tn. lou.ge based on ssc
Marks, HSC Marks, UG Marks, Sports
(collegelevel/ statelevel/ Natioanl revel/internationar reveli.
_-
choose the appropriate Mt ergorithm to solve the above problem
B - with proper argument and3uslification.
A patient takes a lab test and the result comes back positive.
CO2BL3 The test return a correct positive resurt in only 9g% t6I
cases in which the disease is actualry present,
of the
ind
negative result in only 97o/o of the cases in which the
a correct
diseases
is not present. Furthermore, .o0g of the entire population
have this cancer' Does the patient have the cancer
or not.
OR

P^^__ Using the following training set identify the class for the
instance
CO2BL3 given in the test in the table using multinomial narVe bayes
t6I
classification.
Doc Words Class
Training 1 Chinese Beijing Chinese C
2 Chinese Chinese Shanehai C
J Chinese Macao C
4 Tokvo Japal Chinese i
Test 5 Chinese Chinese Chinese ToLyo ?
Japan
C D_erive the expression for logistic regression
COIBL2 for two variable data t6l
distribution. Assume activation to be a sigmoid function.

CO = Course Outcome No
BL = Bloom Taxonomy Level

p^*<_ +/t